aboutsummaryrefslogtreecommitdiffstats
path: root/net/ipv4/ip_gre.c
diff options
context:
space:
mode:
Diffstat (limited to 'net/ipv4/ip_gre.c')
-rw-r--r--net/ipv4/ip_gre.c1516
1 files changed, 316 insertions, 1200 deletions
diff --git a/net/ipv4/ip_gre.c b/net/ipv4/ip_gre.c
index 91d66dbde9c0..ad662e906f7e 100644
--- a/net/ipv4/ip_gre.c
+++ b/net/ipv4/ip_gre.c
@@ -37,7 +37,7 @@
37#include <net/ip.h> 37#include <net/ip.h>
38#include <net/icmp.h> 38#include <net/icmp.h>
39#include <net/protocol.h> 39#include <net/protocol.h>
40#include <net/ipip.h> 40#include <net/ip_tunnels.h>
41#include <net/arp.h> 41#include <net/arp.h>
42#include <net/checksum.h> 42#include <net/checksum.h>
43#include <net/dsfield.h> 43#include <net/dsfield.h>
@@ -108,15 +108,6 @@
108 fatal route to network, even if it were you who configured 108 fatal route to network, even if it were you who configured
109 fatal static route: you are innocent. :-) 109 fatal static route: you are innocent. :-)
110 110
111
112
113 3. Really, ipv4/ipip.c, ipv4/ip_gre.c and ipv6/sit.c contain
114 practically identical code. It would be good to glue them
115 together, but it is not very evident, how to make them modular.
116 sit is integral part of IPv6, ipip and gre are naturally modular.
117 We could extract common parts (hash table, ioctl etc)
118 to a separate module (ip_tunnel.c).
119
120 Alexey Kuznetsov. 111 Alexey Kuznetsov.
121 */ 112 */
122 113
@@ -126,400 +117,135 @@ MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
126 117
127static struct rtnl_link_ops ipgre_link_ops __read_mostly; 118static struct rtnl_link_ops ipgre_link_ops __read_mostly;
128static int ipgre_tunnel_init(struct net_device *dev); 119static int ipgre_tunnel_init(struct net_device *dev);
129static void ipgre_tunnel_setup(struct net_device *dev);
130static int ipgre_tunnel_bind_dev(struct net_device *dev);
131
132/* Fallback tunnel: no source, no destination, no key, no options */
133
134#define HASH_SIZE 16
135 120
136static int ipgre_net_id __read_mostly; 121static int ipgre_net_id __read_mostly;
137struct ipgre_net { 122static int gre_tap_net_id __read_mostly;
138 struct ip_tunnel __rcu *tunnels[4][HASH_SIZE];
139
140 struct net_device *fb_tunnel_dev;
141};
142
143/* Tunnel hash table */
144
145/*
146 4 hash tables:
147
148 3: (remote,local)
149 2: (remote,*)
150 1: (*,local)
151 0: (*,*)
152 123
153 We require exact key match i.e. if a key is present in packet 124static __sum16 check_checksum(struct sk_buff *skb)
154 it will match only tunnel with the same key; if it is not present, 125{
155 it will match only keyless tunnel. 126 __sum16 csum = 0;
156
157 All keysless packets, if not matched configured keyless tunnels
158 will match fallback tunnel.
159 */
160 127
161#define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF) 128 switch (skb->ip_summed) {
129 case CHECKSUM_COMPLETE:
130 csum = csum_fold(skb->csum);
162 131
163#define tunnels_r_l tunnels[3] 132 if (!csum)
164#define tunnels_r tunnels[2] 133 break;
165#define tunnels_l tunnels[1] 134 /* Fall through. */
166#define tunnels_wc tunnels[0]
167 135
168static struct rtnl_link_stats64 *ipgre_get_stats64(struct net_device *dev, 136 case CHECKSUM_NONE:
169 struct rtnl_link_stats64 *tot) 137 skb->csum = 0;
170{ 138 csum = __skb_checksum_complete(skb);
171 int i; 139 skb->ip_summed = CHECKSUM_COMPLETE;
172 140 break;
173 for_each_possible_cpu(i) {
174 const struct pcpu_tstats *tstats = per_cpu_ptr(dev->tstats, i);
175 u64 rx_packets, rx_bytes, tx_packets, tx_bytes;
176 unsigned int start;
177
178 do {
179 start = u64_stats_fetch_begin_bh(&tstats->syncp);
180 rx_packets = tstats->rx_packets;
181 tx_packets = tstats->tx_packets;
182 rx_bytes = tstats->rx_bytes;
183 tx_bytes = tstats->tx_bytes;
184 } while (u64_stats_fetch_retry_bh(&tstats->syncp, start));
185
186 tot->rx_packets += rx_packets;
187 tot->tx_packets += tx_packets;
188 tot->rx_bytes += rx_bytes;
189 tot->tx_bytes += tx_bytes;
190 } 141 }
191 142
192 tot->multicast = dev->stats.multicast; 143 return csum;
193 tot->rx_crc_errors = dev->stats.rx_crc_errors;
194 tot->rx_fifo_errors = dev->stats.rx_fifo_errors;
195 tot->rx_length_errors = dev->stats.rx_length_errors;
196 tot->rx_frame_errors = dev->stats.rx_frame_errors;
197 tot->rx_errors = dev->stats.rx_errors;
198
199 tot->tx_fifo_errors = dev->stats.tx_fifo_errors;
200 tot->tx_carrier_errors = dev->stats.tx_carrier_errors;
201 tot->tx_dropped = dev->stats.tx_dropped;
202 tot->tx_aborted_errors = dev->stats.tx_aborted_errors;
203 tot->tx_errors = dev->stats.tx_errors;
204
205 return tot;
206} 144}
207 145
208/* Does key in tunnel parameters match packet */ 146static int ip_gre_calc_hlen(__be16 o_flags)
209static bool ipgre_key_match(const struct ip_tunnel_parm *p,
210 __be16 flags, __be32 key)
211{ 147{
212 if (p->i_flags & GRE_KEY) { 148 int addend = 4;
213 if (flags & GRE_KEY)
214 return key == p->i_key;
215 else
216 return false; /* key expected, none present */
217 } else
218 return !(flags & GRE_KEY);
219}
220 149
221/* Given src, dst and key, find appropriate for input tunnel. */ 150 if (o_flags&TUNNEL_CSUM)
151 addend += 4;
152 if (o_flags&TUNNEL_KEY)
153 addend += 4;
154 if (o_flags&TUNNEL_SEQ)
155 addend += 4;
156 return addend;
157}
222 158
223static struct ip_tunnel *ipgre_tunnel_lookup(struct net_device *dev, 159static int parse_gre_header(struct sk_buff *skb, struct tnl_ptk_info *tpi,
224 __be32 remote, __be32 local, 160 bool *csum_err, int *hdr_len)
225 __be16 flags, __be32 key,
226 __be16 gre_proto)
227{ 161{
228 struct net *net = dev_net(dev); 162 struct iphdr *iph = ip_hdr(skb);
229 int link = dev->ifindex; 163 struct gre_base_hdr *greh;
230 unsigned int h0 = HASH(remote); 164 __be32 *options;
231 unsigned int h1 = HASH(key);
232 struct ip_tunnel *t, *cand = NULL;
233 struct ipgre_net *ign = net_generic(net, ipgre_net_id);
234 int dev_type = (gre_proto == htons(ETH_P_TEB)) ?
235 ARPHRD_ETHER : ARPHRD_IPGRE;
236 int score, cand_score = 4;
237
238 for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) {
239 if (local != t->parms.iph.saddr ||
240 remote != t->parms.iph.daddr ||
241 !(t->dev->flags & IFF_UP))
242 continue;
243
244 if (!ipgre_key_match(&t->parms, flags, key))
245 continue;
246
247 if (t->dev->type != ARPHRD_IPGRE &&
248 t->dev->type != dev_type)
249 continue;
250
251 score = 0;
252 if (t->parms.link != link)
253 score |= 1;
254 if (t->dev->type != dev_type)
255 score |= 2;
256 if (score == 0)
257 return t;
258
259 if (score < cand_score) {
260 cand = t;
261 cand_score = score;
262 }
263 }
264
265 for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) {
266 if (remote != t->parms.iph.daddr ||
267 !(t->dev->flags & IFF_UP))
268 continue;
269
270 if (!ipgre_key_match(&t->parms, flags, key))
271 continue;
272
273 if (t->dev->type != ARPHRD_IPGRE &&
274 t->dev->type != dev_type)
275 continue;
276
277 score = 0;
278 if (t->parms.link != link)
279 score |= 1;
280 if (t->dev->type != dev_type)
281 score |= 2;
282 if (score == 0)
283 return t;
284
285 if (score < cand_score) {
286 cand = t;
287 cand_score = score;
288 }
289 }
290 165
291 for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) { 166 if (unlikely(!pskb_may_pull(skb, sizeof(struct gre_base_hdr))))
292 if ((local != t->parms.iph.saddr && 167 return -EINVAL;
293 (local != t->parms.iph.daddr ||
294 !ipv4_is_multicast(local))) ||
295 !(t->dev->flags & IFF_UP))
296 continue;
297
298 if (!ipgre_key_match(&t->parms, flags, key))
299 continue;
300
301 if (t->dev->type != ARPHRD_IPGRE &&
302 t->dev->type != dev_type)
303 continue;
304
305 score = 0;
306 if (t->parms.link != link)
307 score |= 1;
308 if (t->dev->type != dev_type)
309 score |= 2;
310 if (score == 0)
311 return t;
312
313 if (score < cand_score) {
314 cand = t;
315 cand_score = score;
316 }
317 }
318 168
319 for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) { 169 greh = (struct gre_base_hdr *)((u8 *)iph + (iph->ihl << 2));
320 if (t->parms.i_key != key || 170 if (unlikely(greh->flags & (GRE_VERSION | GRE_ROUTING)))
321 !(t->dev->flags & IFF_UP)) 171 return -EINVAL;
322 continue;
323
324 if (t->dev->type != ARPHRD_IPGRE &&
325 t->dev->type != dev_type)
326 continue;
327
328 score = 0;
329 if (t->parms.link != link)
330 score |= 1;
331 if (t->dev->type != dev_type)
332 score |= 2;
333 if (score == 0)
334 return t;
335
336 if (score < cand_score) {
337 cand = t;
338 cand_score = score;
339 }
340 }
341 172
342 if (cand != NULL) 173 tpi->flags = gre_flags_to_tnl_flags(greh->flags);
343 return cand; 174 *hdr_len = ip_gre_calc_hlen(tpi->flags);
344 175
345 dev = ign->fb_tunnel_dev; 176 if (!pskb_may_pull(skb, *hdr_len))
346 if (dev->flags & IFF_UP) 177 return -EINVAL;
347 return netdev_priv(dev);
348 178
349 return NULL; 179 tpi->proto = greh->protocol;
350}
351 180
352static struct ip_tunnel __rcu **__ipgre_bucket(struct ipgre_net *ign, 181 options = (__be32 *)(greh + 1);
353 struct ip_tunnel_parm *parms) 182 if (greh->flags & GRE_CSUM) {
354{ 183 if (check_checksum(skb)) {
355 __be32 remote = parms->iph.daddr; 184 *csum_err = true;
356 __be32 local = parms->iph.saddr; 185 return -EINVAL;
357 __be32 key = parms->i_key; 186 }
358 unsigned int h = HASH(key); 187 options++;
359 int prio = 0;
360
361 if (local)
362 prio |= 1;
363 if (remote && !ipv4_is_multicast(remote)) {
364 prio |= 2;
365 h ^= HASH(remote);
366 } 188 }
367 189
368 return &ign->tunnels[prio][h]; 190 if (greh->flags & GRE_KEY) {
369} 191 tpi->key = *options;
370 192 options++;
371static inline struct ip_tunnel __rcu **ipgre_bucket(struct ipgre_net *ign, 193 } else
372 struct ip_tunnel *t) 194 tpi->key = 0;
373{
374 return __ipgre_bucket(ign, &t->parms);
375}
376
377static void ipgre_tunnel_link(struct ipgre_net *ign, struct ip_tunnel *t)
378{
379 struct ip_tunnel __rcu **tp = ipgre_bucket(ign, t);
380 195
381 rcu_assign_pointer(t->next, rtnl_dereference(*tp)); 196 if (unlikely(greh->flags & GRE_SEQ)) {
382 rcu_assign_pointer(*tp, t); 197 tpi->seq = *options;
383} 198 options++;
199 } else
200 tpi->seq = 0;
384 201
385static void ipgre_tunnel_unlink(struct ipgre_net *ign, struct ip_tunnel *t) 202 /* WCCP version 1 and 2 protocol decoding.
386{ 203 * - Change protocol to IP
387 struct ip_tunnel __rcu **tp; 204 * - When dealing with WCCPv2, Skip extra 4 bytes in GRE header
388 struct ip_tunnel *iter; 205 */
389 206 if (greh->flags == 0 && tpi->proto == htons(ETH_P_WCCP)) {
390 for (tp = ipgre_bucket(ign, t); 207 tpi->proto = htons(ETH_P_IP);
391 (iter = rtnl_dereference(*tp)) != NULL; 208 if ((*(u8 *)options & 0xF0) != 0x40) {
392 tp = &iter->next) { 209 *hdr_len += 4;
393 if (t == iter) { 210 if (!pskb_may_pull(skb, *hdr_len))
394 rcu_assign_pointer(*tp, t->next); 211 return -EINVAL;
395 break;
396 } 212 }
397 } 213 }
398}
399
400static struct ip_tunnel *ipgre_tunnel_find(struct net *net,
401 struct ip_tunnel_parm *parms,
402 int type)
403{
404 __be32 remote = parms->iph.daddr;
405 __be32 local = parms->iph.saddr;
406 __be32 key = parms->i_key;
407 int link = parms->link;
408 struct ip_tunnel *t;
409 struct ip_tunnel __rcu **tp;
410 struct ipgre_net *ign = net_generic(net, ipgre_net_id);
411
412 for (tp = __ipgre_bucket(ign, parms);
413 (t = rtnl_dereference(*tp)) != NULL;
414 tp = &t->next)
415 if (local == t->parms.iph.saddr &&
416 remote == t->parms.iph.daddr &&
417 key == t->parms.i_key &&
418 link == t->parms.link &&
419 type == t->dev->type)
420 break;
421
422 return t;
423}
424
425static struct ip_tunnel *ipgre_tunnel_locate(struct net *net,
426 struct ip_tunnel_parm *parms, int create)
427{
428 struct ip_tunnel *t, *nt;
429 struct net_device *dev;
430 char name[IFNAMSIZ];
431 struct ipgre_net *ign = net_generic(net, ipgre_net_id);
432
433 t = ipgre_tunnel_find(net, parms, ARPHRD_IPGRE);
434 if (t || !create)
435 return t;
436
437 if (parms->name[0])
438 strlcpy(name, parms->name, IFNAMSIZ);
439 else
440 strcpy(name, "gre%d");
441
442 dev = alloc_netdev(sizeof(*t), name, ipgre_tunnel_setup);
443 if (!dev)
444 return NULL;
445
446 dev_net_set(dev, net);
447
448 nt = netdev_priv(dev);
449 nt->parms = *parms;
450 dev->rtnl_link_ops = &ipgre_link_ops;
451 214
452 dev->mtu = ipgre_tunnel_bind_dev(dev); 215 return 0;
453
454 if (register_netdevice(dev) < 0)
455 goto failed_free;
456
457 /* Can use a lockless transmit, unless we generate output sequences */
458 if (!(nt->parms.o_flags & GRE_SEQ))
459 dev->features |= NETIF_F_LLTX;
460
461 dev_hold(dev);
462 ipgre_tunnel_link(ign, nt);
463 return nt;
464
465failed_free:
466 free_netdev(dev);
467 return NULL;
468}
469
470static void ipgre_tunnel_uninit(struct net_device *dev)
471{
472 struct net *net = dev_net(dev);
473 struct ipgre_net *ign = net_generic(net, ipgre_net_id);
474
475 ipgre_tunnel_unlink(ign, netdev_priv(dev));
476 dev_put(dev);
477} 216}
478 217
479
480static void ipgre_err(struct sk_buff *skb, u32 info) 218static void ipgre_err(struct sk_buff *skb, u32 info)
481{ 219{
482 220
483/* All the routers (except for Linux) return only 221 /* All the routers (except for Linux) return only
484 8 bytes of packet payload. It means, that precise relaying of 222 8 bytes of packet payload. It means, that precise relaying of
485 ICMP in the real Internet is absolutely infeasible. 223 ICMP in the real Internet is absolutely infeasible.
486 224
487 Moreover, Cisco "wise men" put GRE key to the third word 225 Moreover, Cisco "wise men" put GRE key to the third word
488 in GRE header. It makes impossible maintaining even soft state for keyed 226 in GRE header. It makes impossible maintaining even soft
489 GRE tunnels with enabled checksum. Tell them "thank you". 227 state for keyed GRE tunnels with enabled checksum. Tell
490 228 them "thank you".
491 Well, I wonder, rfc1812 was written by Cisco employee,
492 what the hell these idiots break standards established
493 by themselves???
494 */
495 229
230 Well, I wonder, rfc1812 was written by Cisco employee,
231 what the hell these idiots break standards established
232 by themselves???
233 */
234 struct net *net = dev_net(skb->dev);
235 struct ip_tunnel_net *itn;
496 const struct iphdr *iph = (const struct iphdr *)skb->data; 236 const struct iphdr *iph = (const struct iphdr *)skb->data;
497 __be16 *p = (__be16 *)(skb->data+(iph->ihl<<2));
498 int grehlen = (iph->ihl<<2) + 4;
499 const int type = icmp_hdr(skb)->type; 237 const int type = icmp_hdr(skb)->type;
500 const int code = icmp_hdr(skb)->code; 238 const int code = icmp_hdr(skb)->code;
501 struct ip_tunnel *t; 239 struct ip_tunnel *t;
502 __be16 flags; 240 struct tnl_ptk_info tpi;
503 __be32 key = 0; 241 int hdr_len;
242 bool csum_err = false;
504 243
505 flags = p[0]; 244 if (parse_gre_header(skb, &tpi, &csum_err, &hdr_len)) {
506 if (flags&(GRE_CSUM|GRE_KEY|GRE_SEQ|GRE_ROUTING|GRE_VERSION)) { 245 if (!csum_err) /* ignore csum errors. */
507 if (flags&(GRE_VERSION|GRE_ROUTING))
508 return; 246 return;
509 if (flags&GRE_KEY) {
510 grehlen += 4;
511 if (flags&GRE_CSUM)
512 grehlen += 4;
513 }
514 } 247 }
515 248
516 /* If only 8 bytes returned, keyed message will be dropped here */
517 if (skb_headlen(skb) < grehlen)
518 return;
519
520 if (flags & GRE_KEY)
521 key = *(((__be32 *)p) + (grehlen / 4) - 1);
522
523 switch (type) { 249 switch (type) {
524 default: 250 default:
525 case ICMP_PARAMETERPROB: 251 case ICMP_PARAMETERPROB:
@@ -548,8 +274,13 @@ static void ipgre_err(struct sk_buff *skb, u32 info)
548 break; 274 break;
549 } 275 }
550 276
551 t = ipgre_tunnel_lookup(skb->dev, iph->daddr, iph->saddr, 277 if (tpi.proto == htons(ETH_P_TEB))
552 flags, key, p[1]); 278 itn = net_generic(net, gre_tap_net_id);
279 else
280 itn = net_generic(net, ipgre_net_id);
281
282 t = ip_tunnel_lookup(itn, skb->dev->ifindex, tpi.flags,
283 iph->daddr, iph->saddr, tpi.key);
553 284
554 if (t == NULL) 285 if (t == NULL)
555 return; 286 return;
@@ -578,158 +309,33 @@ static void ipgre_err(struct sk_buff *skb, u32 info)
578 t->err_time = jiffies; 309 t->err_time = jiffies;
579} 310}
580 311
581static inline u8
582ipgre_ecn_encapsulate(u8 tos, const struct iphdr *old_iph, struct sk_buff *skb)
583{
584 u8 inner = 0;
585 if (skb->protocol == htons(ETH_P_IP))
586 inner = old_iph->tos;
587 else if (skb->protocol == htons(ETH_P_IPV6))
588 inner = ipv6_get_dsfield((const struct ipv6hdr *)old_iph);
589 return INET_ECN_encapsulate(tos, inner);
590}
591
592static int ipgre_rcv(struct sk_buff *skb) 312static int ipgre_rcv(struct sk_buff *skb)
593{ 313{
314 struct net *net = dev_net(skb->dev);
315 struct ip_tunnel_net *itn;
594 const struct iphdr *iph; 316 const struct iphdr *iph;
595 u8 *h;
596 __be16 flags;
597 __sum16 csum = 0;
598 __be32 key = 0;
599 u32 seqno = 0;
600 struct ip_tunnel *tunnel; 317 struct ip_tunnel *tunnel;
601 int offset = 4; 318 struct tnl_ptk_info tpi;
602 __be16 gre_proto; 319 int hdr_len;
603 int err; 320 bool csum_err = false;
604 321
605 if (!pskb_may_pull(skb, 16)) 322 if (parse_gre_header(skb, &tpi, &csum_err, &hdr_len) < 0)
606 goto drop; 323 goto drop;
607 324
608 iph = ip_hdr(skb); 325 if (tpi.proto == htons(ETH_P_TEB))
609 h = skb->data; 326 itn = net_generic(net, gre_tap_net_id);
610 flags = *(__be16 *)h; 327 else
611 328 itn = net_generic(net, ipgre_net_id);
612 if (flags&(GRE_CSUM|GRE_KEY|GRE_ROUTING|GRE_SEQ|GRE_VERSION)) {
613 /* - Version must be 0.
614 - We do not support routing headers.
615 */
616 if (flags&(GRE_VERSION|GRE_ROUTING))
617 goto drop;
618
619 if (flags&GRE_CSUM) {
620 switch (skb->ip_summed) {
621 case CHECKSUM_COMPLETE:
622 csum = csum_fold(skb->csum);
623 if (!csum)
624 break;
625 /* fall through */
626 case CHECKSUM_NONE:
627 skb->csum = 0;
628 csum = __skb_checksum_complete(skb);
629 skb->ip_summed = CHECKSUM_COMPLETE;
630 }
631 offset += 4;
632 }
633 if (flags&GRE_KEY) {
634 key = *(__be32 *)(h + offset);
635 offset += 4;
636 }
637 if (flags&GRE_SEQ) {
638 seqno = ntohl(*(__be32 *)(h + offset));
639 offset += 4;
640 }
641 }
642 329
643 gre_proto = *(__be16 *)(h + 2); 330 iph = ip_hdr(skb);
331 tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, tpi.flags,
332 iph->saddr, iph->daddr, tpi.key);
644 333
645 tunnel = ipgre_tunnel_lookup(skb->dev,
646 iph->saddr, iph->daddr, flags, key,
647 gre_proto);
648 if (tunnel) { 334 if (tunnel) {
649 struct pcpu_tstats *tstats; 335 ip_tunnel_rcv(tunnel, skb, &tpi, log_ecn_error);
650
651 secpath_reset(skb);
652
653 skb->protocol = gre_proto;
654 /* WCCP version 1 and 2 protocol decoding.
655 * - Change protocol to IP
656 * - When dealing with WCCPv2, Skip extra 4 bytes in GRE header
657 */
658 if (flags == 0 && gre_proto == htons(ETH_P_WCCP)) {
659 skb->protocol = htons(ETH_P_IP);
660 if ((*(h + offset) & 0xF0) != 0x40)
661 offset += 4;
662 }
663
664 skb->mac_header = skb->network_header;
665 __pskb_pull(skb, offset);
666 skb_postpull_rcsum(skb, skb_transport_header(skb), offset);
667 skb->pkt_type = PACKET_HOST;
668#ifdef CONFIG_NET_IPGRE_BROADCAST
669 if (ipv4_is_multicast(iph->daddr)) {
670 /* Looped back packet, drop it! */
671 if (rt_is_output_route(skb_rtable(skb)))
672 goto drop;
673 tunnel->dev->stats.multicast++;
674 skb->pkt_type = PACKET_BROADCAST;
675 }
676#endif
677
678 if (((flags&GRE_CSUM) && csum) ||
679 (!(flags&GRE_CSUM) && tunnel->parms.i_flags&GRE_CSUM)) {
680 tunnel->dev->stats.rx_crc_errors++;
681 tunnel->dev->stats.rx_errors++;
682 goto drop;
683 }
684 if (tunnel->parms.i_flags&GRE_SEQ) {
685 if (!(flags&GRE_SEQ) ||
686 (tunnel->i_seqno && (s32)(seqno - tunnel->i_seqno) < 0)) {
687 tunnel->dev->stats.rx_fifo_errors++;
688 tunnel->dev->stats.rx_errors++;
689 goto drop;
690 }
691 tunnel->i_seqno = seqno + 1;
692 }
693
694 /* Warning: All skb pointers will be invalidated! */
695 if (tunnel->dev->type == ARPHRD_ETHER) {
696 if (!pskb_may_pull(skb, ETH_HLEN)) {
697 tunnel->dev->stats.rx_length_errors++;
698 tunnel->dev->stats.rx_errors++;
699 goto drop;
700 }
701
702 iph = ip_hdr(skb);
703 skb->protocol = eth_type_trans(skb, tunnel->dev);
704 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
705 }
706
707 __skb_tunnel_rx(skb, tunnel->dev);
708
709 skb_reset_network_header(skb);
710 err = IP_ECN_decapsulate(iph, skb);
711 if (unlikely(err)) {
712 if (log_ecn_error)
713 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
714 &iph->saddr, iph->tos);
715 if (err > 1) {
716 ++tunnel->dev->stats.rx_frame_errors;
717 ++tunnel->dev->stats.rx_errors;
718 goto drop;
719 }
720 }
721
722 tstats = this_cpu_ptr(tunnel->dev->tstats);
723 u64_stats_update_begin(&tstats->syncp);
724 tstats->rx_packets++;
725 tstats->rx_bytes += skb->len;
726 u64_stats_update_end(&tstats->syncp);
727
728 gro_cells_receive(&tunnel->gro_cells, skb);
729 return 0; 336 return 0;
730 } 337 }
731 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0); 338 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
732
733drop: 339drop:
734 kfree_skb(skb); 340 kfree_skb(skb);
735 return 0; 341 return 0;
@@ -746,7 +352,7 @@ static struct sk_buff *handle_offloads(struct ip_tunnel *tunnel, struct sk_buff
746 skb_shinfo(skb)->gso_type |= SKB_GSO_GRE; 352 skb_shinfo(skb)->gso_type |= SKB_GSO_GRE;
747 return skb; 353 return skb;
748 } else if (skb->ip_summed == CHECKSUM_PARTIAL && 354 } else if (skb->ip_summed == CHECKSUM_PARTIAL &&
749 tunnel->parms.o_flags&GRE_CSUM) { 355 tunnel->parms.o_flags&TUNNEL_CSUM) {
750 err = skb_checksum_help(skb); 356 err = skb_checksum_help(skb);
751 if (unlikely(err)) 357 if (unlikely(err))
752 goto error; 358 goto error;
@@ -760,494 +366,157 @@ error:
760 return ERR_PTR(err); 366 return ERR_PTR(err);
761} 367}
762 368
763static netdev_tx_t ipgre_tunnel_xmit(struct sk_buff *skb, struct net_device *dev) 369static struct sk_buff *gre_build_header(struct sk_buff *skb,
370 const struct tnl_ptk_info *tpi,
371 int hdr_len)
764{ 372{
765 struct pcpu_tstats *tstats = this_cpu_ptr(dev->tstats); 373 struct gre_base_hdr *greh;
766 struct ip_tunnel *tunnel = netdev_priv(dev);
767 const struct iphdr *old_iph;
768 const struct iphdr *tiph;
769 struct flowi4 fl4;
770 u8 tos;
771 __be16 df;
772 struct rtable *rt; /* Route to the other host */
773 struct net_device *tdev; /* Device to other host */
774 struct iphdr *iph; /* Our new IP header */
775 unsigned int max_headroom; /* The extra header space needed */
776 int gre_hlen;
777 __be32 dst;
778 int mtu;
779 u8 ttl;
780 int err;
781 int pkt_len;
782
783 skb = handle_offloads(tunnel, skb);
784 if (IS_ERR(skb)) {
785 dev->stats.tx_dropped++;
786 return NETDEV_TX_OK;
787 }
788 374
789 if (!skb->encapsulation) { 375 skb_push(skb, hdr_len);
790 skb_reset_inner_headers(skb);
791 skb->encapsulation = 1;
792 }
793 376
794 old_iph = ip_hdr(skb); 377 greh = (struct gre_base_hdr *)skb->data;
378 greh->flags = tnl_flags_to_gre_flags(tpi->flags);
379 greh->protocol = tpi->proto;
795 380
796 if (dev->type == ARPHRD_ETHER) 381 if (tpi->flags&(TUNNEL_KEY|TUNNEL_CSUM|TUNNEL_SEQ)) {
797 IPCB(skb)->flags = 0; 382 __be32 *ptr = (__be32 *)(((u8 *)greh) + hdr_len - 4);
798 383
799 if (dev->header_ops && dev->type == ARPHRD_IPGRE) { 384 if (tpi->flags&TUNNEL_SEQ) {
800 gre_hlen = 0; 385 *ptr = tpi->seq;
801 tiph = (const struct iphdr *)skb->data; 386 ptr--;
802 } else {
803 gre_hlen = tunnel->hlen;
804 tiph = &tunnel->parms.iph;
805 }
806
807 if ((dst = tiph->daddr) == 0) {
808 /* NBMA tunnel */
809
810 if (skb_dst(skb) == NULL) {
811 dev->stats.tx_fifo_errors++;
812 goto tx_error;
813 } 387 }
814 388 if (tpi->flags&TUNNEL_KEY) {
815 if (skb->protocol == htons(ETH_P_IP)) { 389 *ptr = tpi->key;
816 rt = skb_rtable(skb); 390 ptr--;
817 dst = rt_nexthop(rt, old_iph->daddr);
818 } 391 }
819#if IS_ENABLED(CONFIG_IPV6) 392 if (tpi->flags&TUNNEL_CSUM &&
820 else if (skb->protocol == htons(ETH_P_IPV6)) { 393 !(skb_shinfo(skb)->gso_type & SKB_GSO_GRE)) {
821 const struct in6_addr *addr6; 394 *(__sum16 *)ptr = 0;
822 struct neighbour *neigh; 395 *(__sum16 *)ptr = csum_fold(skb_checksum(skb, 0,
823 bool do_tx_error_icmp; 396 skb->len, 0));
824 int addr_type;
825
826 neigh = dst_neigh_lookup(skb_dst(skb), &ipv6_hdr(skb)->daddr);
827 if (neigh == NULL)
828 goto tx_error;
829
830 addr6 = (const struct in6_addr *)&neigh->primary_key;
831 addr_type = ipv6_addr_type(addr6);
832
833 if (addr_type == IPV6_ADDR_ANY) {
834 addr6 = &ipv6_hdr(skb)->daddr;
835 addr_type = ipv6_addr_type(addr6);
836 }
837
838 if ((addr_type & IPV6_ADDR_COMPATv4) == 0)
839 do_tx_error_icmp = true;
840 else {
841 do_tx_error_icmp = false;
842 dst = addr6->s6_addr32[3];
843 }
844 neigh_release(neigh);
845 if (do_tx_error_icmp)
846 goto tx_error_icmp;
847 } 397 }
848#endif
849 else
850 goto tx_error;
851 } 398 }
852 399
853 ttl = tiph->ttl; 400 return skb;
854 tos = tiph->tos; 401}
855 if (tos & 0x1) {
856 tos &= ~0x1;
857 if (skb->protocol == htons(ETH_P_IP))
858 tos = old_iph->tos;
859 else if (skb->protocol == htons(ETH_P_IPV6))
860 tos = ipv6_get_dsfield((const struct ipv6hdr *)old_iph);
861 }
862 402
863 rt = ip_route_output_gre(dev_net(dev), &fl4, dst, tiph->saddr, 403static void __gre_xmit(struct sk_buff *skb, struct net_device *dev,
864 tunnel->parms.o_key, RT_TOS(tos), 404 const struct iphdr *tnl_params,
865 tunnel->parms.link); 405 __be16 proto)
866 if (IS_ERR(rt)) { 406{
867 dev->stats.tx_carrier_errors++; 407 struct ip_tunnel *tunnel = netdev_priv(dev);
868 goto tx_error; 408 struct tnl_ptk_info tpi;
869 }
870 tdev = rt->dst.dev;
871 409
872 if (tdev == dev) { 410 if (likely(!skb->encapsulation)) {
873 ip_rt_put(rt); 411 skb_reset_inner_headers(skb);
874 dev->stats.collisions++; 412 skb->encapsulation = 1;
875 goto tx_error;
876 } 413 }
877 414
878 df = tiph->frag_off; 415 tpi.flags = tunnel->parms.o_flags;
879 if (df) 416 tpi.proto = proto;
880 mtu = dst_mtu(&rt->dst) - dev->hard_header_len - tunnel->hlen; 417 tpi.key = tunnel->parms.o_key;
881 else 418 if (tunnel->parms.o_flags & TUNNEL_SEQ)
882 mtu = skb_dst(skb) ? dst_mtu(skb_dst(skb)) : dev->mtu; 419 tunnel->o_seqno++;
883 420 tpi.seq = htonl(tunnel->o_seqno);
884 if (skb_dst(skb))
885 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
886
887 if (skb->protocol == htons(ETH_P_IP)) {
888 df |= (old_iph->frag_off&htons(IP_DF));
889 421
890 if (!skb_is_gso(skb) && 422 /* Push GRE header. */
891 (old_iph->frag_off&htons(IP_DF)) && 423 skb = gre_build_header(skb, &tpi, tunnel->hlen);
892 mtu < ntohs(old_iph->tot_len)) { 424 if (unlikely(!skb)) {
893 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, htonl(mtu)); 425 dev->stats.tx_dropped++;
894 ip_rt_put(rt); 426 return;
895 goto tx_error;
896 }
897 } 427 }
898#if IS_ENABLED(CONFIG_IPV6)
899 else if (skb->protocol == htons(ETH_P_IPV6)) {
900 struct rt6_info *rt6 = (struct rt6_info *)skb_dst(skb);
901
902 if (rt6 && mtu < dst_mtu(skb_dst(skb)) && mtu >= IPV6_MIN_MTU) {
903 if ((tunnel->parms.iph.daddr &&
904 !ipv4_is_multicast(tunnel->parms.iph.daddr)) ||
905 rt6->rt6i_dst.plen == 128) {
906 rt6->rt6i_flags |= RTF_MODIFIED;
907 dst_metric_set(skb_dst(skb), RTAX_MTU, mtu);
908 }
909 }
910 428
911 if (!skb_is_gso(skb) && 429 ip_tunnel_xmit(skb, dev, tnl_params);
912 mtu >= IPV6_MIN_MTU && 430}
913 mtu < skb->len - tunnel->hlen + gre_hlen) {
914 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
915 ip_rt_put(rt);
916 goto tx_error;
917 }
918 }
919#endif
920 431
921 if (tunnel->err_count > 0) { 432static netdev_tx_t ipgre_xmit(struct sk_buff *skb,
922 if (time_before(jiffies, 433 struct net_device *dev)
923 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) { 434{
924 tunnel->err_count--; 435 struct ip_tunnel *tunnel = netdev_priv(dev);
436 const struct iphdr *tnl_params;
925 437
926 dst_link_failure(skb); 438 skb = handle_offloads(tunnel, skb);
927 } else 439 if (IS_ERR(skb))
928 tunnel->err_count = 0; 440 goto out;
929 }
930 441
931 max_headroom = LL_RESERVED_SPACE(tdev) + gre_hlen + rt->dst.header_len; 442 if (dev->header_ops) {
932 443 /* Need space for new headers */
933 if (skb_headroom(skb) < max_headroom || skb_shared(skb)|| 444 if (skb_cow_head(skb, dev->needed_headroom -
934 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) { 445 (tunnel->hlen + sizeof(struct iphdr))));
935 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom); 446 goto free_skb;
936 if (max_headroom > dev->needed_headroom)
937 dev->needed_headroom = max_headroom;
938 if (!new_skb) {
939 ip_rt_put(rt);
940 dev->stats.tx_dropped++;
941 dev_kfree_skb(skb);
942 return NETDEV_TX_OK;
943 }
944 if (skb->sk)
945 skb_set_owner_w(new_skb, skb->sk);
946 dev_kfree_skb(skb);
947 skb = new_skb;
948 old_iph = ip_hdr(skb);
949 /* Warning : tiph value might point to freed memory */
950 }
951 447
952 skb_push(skb, gre_hlen); 448 tnl_params = (const struct iphdr *)skb->data;
953 skb_reset_network_header(skb);
954 skb_set_transport_header(skb, sizeof(*iph));
955 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
956 IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED |
957 IPSKB_REROUTED);
958 skb_dst_drop(skb);
959 skb_dst_set(skb, &rt->dst);
960
961 /*
962 * Push down and install the IPIP header.
963 */
964 449
965 iph = ip_hdr(skb); 450 /* Pull skb since ip_tunnel_xmit() needs skb->data pointing
966 iph->version = 4; 451 * to gre header.
967 iph->ihl = sizeof(struct iphdr) >> 2; 452 */
968 iph->frag_off = df; 453 skb_pull(skb, tunnel->hlen + sizeof(struct iphdr));
969 iph->protocol = IPPROTO_GRE; 454 } else {
970 iph->tos = ipgre_ecn_encapsulate(tos, old_iph, skb); 455 if (skb_cow_head(skb, dev->needed_headroom))
971 iph->daddr = fl4.daddr; 456 goto free_skb;
972 iph->saddr = fl4.saddr;
973 iph->ttl = ttl;
974
975 tunnel_ip_select_ident(skb, old_iph, &rt->dst);
976
977 if (ttl == 0) {
978 if (skb->protocol == htons(ETH_P_IP))
979 iph->ttl = old_iph->ttl;
980#if IS_ENABLED(CONFIG_IPV6)
981 else if (skb->protocol == htons(ETH_P_IPV6))
982 iph->ttl = ((const struct ipv6hdr *)old_iph)->hop_limit;
983#endif
984 else
985 iph->ttl = ip4_dst_hoplimit(&rt->dst);
986 }
987
988 ((__be16 *)(iph + 1))[0] = tunnel->parms.o_flags;
989 ((__be16 *)(iph + 1))[1] = (dev->type == ARPHRD_ETHER) ?
990 htons(ETH_P_TEB) : skb->protocol;
991
992 if (tunnel->parms.o_flags&(GRE_KEY|GRE_CSUM|GRE_SEQ)) {
993 __be32 *ptr = (__be32 *)(((u8 *)iph) + tunnel->hlen - 4);
994 457
995 if (tunnel->parms.o_flags&GRE_SEQ) { 458 tnl_params = &tunnel->parms.iph;
996 ++tunnel->o_seqno;
997 *ptr = htonl(tunnel->o_seqno);
998 ptr--;
999 }
1000 if (tunnel->parms.o_flags&GRE_KEY) {
1001 *ptr = tunnel->parms.o_key;
1002 ptr--;
1003 }
1004 /* Skip GRE checksum if skb is getting offloaded. */
1005 if (!(skb_shinfo(skb)->gso_type & SKB_GSO_GRE) &&
1006 (tunnel->parms.o_flags&GRE_CSUM)) {
1007 int offset = skb_transport_offset(skb);
1008
1009 if (skb_has_shared_frag(skb)) {
1010 err = __skb_linearize(skb);
1011 if (err)
1012 goto tx_error;
1013 }
1014
1015 *ptr = 0;
1016 *(__sum16 *)ptr = csum_fold(skb_checksum(skb, offset,
1017 skb->len - offset,
1018 0));
1019 }
1020 } 459 }
1021 460
1022 nf_reset(skb); 461 __gre_xmit(skb, dev, tnl_params, skb->protocol);
1023 462
1024 pkt_len = skb->len - skb_transport_offset(skb);
1025 err = ip_local_out(skb);
1026 if (likely(net_xmit_eval(err) == 0)) {
1027 u64_stats_update_begin(&tstats->syncp);
1028 tstats->tx_bytes += pkt_len;
1029 tstats->tx_packets++;
1030 u64_stats_update_end(&tstats->syncp);
1031 } else {
1032 dev->stats.tx_errors++;
1033 dev->stats.tx_aborted_errors++;
1034 }
1035 return NETDEV_TX_OK; 463 return NETDEV_TX_OK;
1036 464
1037#if IS_ENABLED(CONFIG_IPV6) 465free_skb:
1038tx_error_icmp:
1039 dst_link_failure(skb);
1040#endif
1041tx_error:
1042 dev->stats.tx_errors++;
1043 dev_kfree_skb(skb); 466 dev_kfree_skb(skb);
467out:
468 dev->stats.tx_dropped++;
1044 return NETDEV_TX_OK; 469 return NETDEV_TX_OK;
1045} 470}
1046 471
1047static int ipgre_tunnel_bind_dev(struct net_device *dev) 472static netdev_tx_t gre_tap_xmit(struct sk_buff *skb,
473 struct net_device *dev)
1048{ 474{
1049 struct net_device *tdev = NULL; 475 struct ip_tunnel *tunnel = netdev_priv(dev);
1050 struct ip_tunnel *tunnel;
1051 const struct iphdr *iph;
1052 int hlen = LL_MAX_HEADER;
1053 int mtu = ETH_DATA_LEN;
1054 int addend = sizeof(struct iphdr) + 4;
1055
1056 tunnel = netdev_priv(dev);
1057 iph = &tunnel->parms.iph;
1058
1059 /* Guess output device to choose reasonable mtu and needed_headroom */
1060
1061 if (iph->daddr) {
1062 struct flowi4 fl4;
1063 struct rtable *rt;
1064
1065 rt = ip_route_output_gre(dev_net(dev), &fl4,
1066 iph->daddr, iph->saddr,
1067 tunnel->parms.o_key,
1068 RT_TOS(iph->tos),
1069 tunnel->parms.link);
1070 if (!IS_ERR(rt)) {
1071 tdev = rt->dst.dev;
1072 ip_rt_put(rt);
1073 }
1074
1075 if (dev->type != ARPHRD_ETHER)
1076 dev->flags |= IFF_POINTOPOINT;
1077 }
1078 476
1079 if (!tdev && tunnel->parms.link) 477 skb = handle_offloads(tunnel, skb);
1080 tdev = __dev_get_by_index(dev_net(dev), tunnel->parms.link); 478 if (IS_ERR(skb))
479 goto out;
1081 480
1082 if (tdev) { 481 if (skb_cow_head(skb, dev->needed_headroom))
1083 hlen = tdev->hard_header_len + tdev->needed_headroom; 482 goto free_skb;
1084 mtu = tdev->mtu;
1085 }
1086 dev->iflink = tunnel->parms.link;
1087
1088 /* Precalculate GRE options length */
1089 if (tunnel->parms.o_flags&(GRE_CSUM|GRE_KEY|GRE_SEQ)) {
1090 if (tunnel->parms.o_flags&GRE_CSUM)
1091 addend += 4;
1092 if (tunnel->parms.o_flags&GRE_KEY)
1093 addend += 4;
1094 if (tunnel->parms.o_flags&GRE_SEQ)
1095 addend += 4;
1096 }
1097 dev->needed_headroom = addend + hlen;
1098 mtu -= dev->hard_header_len + addend;
1099 483
1100 if (mtu < 68) 484 __gre_xmit(skb, dev, &tunnel->parms.iph, htons(ETH_P_TEB));
1101 mtu = 68;
1102 485
1103 tunnel->hlen = addend; 486 return NETDEV_TX_OK;
1104 /* TCP offload with GRE SEQ is not supported. */
1105 if (!(tunnel->parms.o_flags & GRE_SEQ)) {
1106 dev->features |= NETIF_F_GSO_SOFTWARE;
1107 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1108 }
1109 487
1110 return mtu; 488free_skb:
489 dev_kfree_skb(skb);
490out:
491 dev->stats.tx_dropped++;
492 return NETDEV_TX_OK;
1111} 493}
1112 494
1113static int 495static int ipgre_tunnel_ioctl(struct net_device *dev,
1114ipgre_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd) 496 struct ifreq *ifr, int cmd)
1115{ 497{
1116 int err = 0; 498 int err = 0;
1117 struct ip_tunnel_parm p; 499 struct ip_tunnel_parm p;
1118 struct ip_tunnel *t;
1119 struct net *net = dev_net(dev);
1120 struct ipgre_net *ign = net_generic(net, ipgre_net_id);
1121
1122 switch (cmd) {
1123 case SIOCGETTUNNEL:
1124 t = NULL;
1125 if (dev == ign->fb_tunnel_dev) {
1126 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1127 err = -EFAULT;
1128 break;
1129 }
1130 t = ipgre_tunnel_locate(net, &p, 0);
1131 }
1132 if (t == NULL)
1133 t = netdev_priv(dev);
1134 memcpy(&p, &t->parms, sizeof(p));
1135 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1136 err = -EFAULT;
1137 break;
1138
1139 case SIOCADDTUNNEL:
1140 case SIOCCHGTUNNEL:
1141 err = -EPERM;
1142 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1143 goto done;
1144
1145 err = -EFAULT;
1146 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1147 goto done;
1148
1149 err = -EINVAL;
1150 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_GRE ||
1151 p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)) ||
1152 ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING)))
1153 goto done;
1154 if (p.iph.ttl)
1155 p.iph.frag_off |= htons(IP_DF);
1156
1157 if (!(p.i_flags&GRE_KEY))
1158 p.i_key = 0;
1159 if (!(p.o_flags&GRE_KEY))
1160 p.o_key = 0;
1161
1162 t = ipgre_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
1163
1164 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1165 if (t != NULL) {
1166 if (t->dev != dev) {
1167 err = -EEXIST;
1168 break;
1169 }
1170 } else {
1171 unsigned int nflags = 0;
1172
1173 t = netdev_priv(dev);
1174
1175 if (ipv4_is_multicast(p.iph.daddr))
1176 nflags = IFF_BROADCAST;
1177 else if (p.iph.daddr)
1178 nflags = IFF_POINTOPOINT;
1179
1180 if ((dev->flags^nflags)&(IFF_POINTOPOINT|IFF_BROADCAST)) {
1181 err = -EINVAL;
1182 break;
1183 }
1184 ipgre_tunnel_unlink(ign, t);
1185 synchronize_net();
1186 t->parms.iph.saddr = p.iph.saddr;
1187 t->parms.iph.daddr = p.iph.daddr;
1188 t->parms.i_key = p.i_key;
1189 t->parms.o_key = p.o_key;
1190 memcpy(dev->dev_addr, &p.iph.saddr, 4);
1191 memcpy(dev->broadcast, &p.iph.daddr, 4);
1192 ipgre_tunnel_link(ign, t);
1193 netdev_state_change(dev);
1194 }
1195 }
1196
1197 if (t) {
1198 err = 0;
1199 if (cmd == SIOCCHGTUNNEL) {
1200 t->parms.iph.ttl = p.iph.ttl;
1201 t->parms.iph.tos = p.iph.tos;
1202 t->parms.iph.frag_off = p.iph.frag_off;
1203 if (t->parms.link != p.link) {
1204 t->parms.link = p.link;
1205 dev->mtu = ipgre_tunnel_bind_dev(dev);
1206 netdev_state_change(dev);
1207 }
1208 }
1209 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1210 err = -EFAULT;
1211 } else
1212 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1213 break;
1214
1215 case SIOCDELTUNNEL:
1216 err = -EPERM;
1217 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1218 goto done;
1219
1220 if (dev == ign->fb_tunnel_dev) {
1221 err = -EFAULT;
1222 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1223 goto done;
1224 err = -ENOENT;
1225 if ((t = ipgre_tunnel_locate(net, &p, 0)) == NULL)
1226 goto done;
1227 err = -EPERM;
1228 if (t == netdev_priv(ign->fb_tunnel_dev))
1229 goto done;
1230 dev = t->dev;
1231 }
1232 unregister_netdevice(dev);
1233 err = 0;
1234 break;
1235 500
1236 default: 501 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1237 err = -EINVAL; 502 return -EFAULT;
503 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_GRE ||
504 p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)) ||
505 ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))) {
506 return -EINVAL;
1238 } 507 }
508 p.i_flags = gre_flags_to_tnl_flags(p.i_flags);
509 p.o_flags = gre_flags_to_tnl_flags(p.o_flags);
1239 510
1240done: 511 err = ip_tunnel_ioctl(dev, &p, cmd);
1241 return err; 512 if (err)
1242} 513 return err;
1243 514
1244static int ipgre_tunnel_change_mtu(struct net_device *dev, int new_mtu) 515 p.i_flags = tnl_flags_to_gre_flags(p.i_flags);
1245{ 516 p.o_flags = tnl_flags_to_gre_flags(p.o_flags);
1246 struct ip_tunnel *tunnel = netdev_priv(dev); 517
1247 if (new_mtu < 68 || 518 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1248 new_mtu > 0xFFF8 - dev->hard_header_len - tunnel->hlen) 519 return -EFAULT;
1249 return -EINVAL;
1250 dev->mtu = new_mtu;
1251 return 0; 520 return 0;
1252} 521}
1253 522
@@ -1277,25 +546,23 @@ static int ipgre_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1277 ... 546 ...
1278 ftp fec0:6666:6666::193.233.7.65 547 ftp fec0:6666:6666::193.233.7.65
1279 ... 548 ...
1280
1281 */ 549 */
1282
1283static int ipgre_header(struct sk_buff *skb, struct net_device *dev, 550static int ipgre_header(struct sk_buff *skb, struct net_device *dev,
1284 unsigned short type, 551 unsigned short type,
1285 const void *daddr, const void *saddr, unsigned int len) 552 const void *daddr, const void *saddr, unsigned int len)
1286{ 553{
1287 struct ip_tunnel *t = netdev_priv(dev); 554 struct ip_tunnel *t = netdev_priv(dev);
1288 struct iphdr *iph = (struct iphdr *)skb_push(skb, t->hlen); 555 struct iphdr *iph;
1289 __be16 *p = (__be16 *)(iph+1); 556 struct gre_base_hdr *greh;
1290 557
1291 memcpy(iph, &t->parms.iph, sizeof(struct iphdr)); 558 iph = (struct iphdr *)skb_push(skb, t->hlen + sizeof(*iph));
1292 p[0] = t->parms.o_flags; 559 greh = (struct gre_base_hdr *)(iph+1);
1293 p[1] = htons(type); 560 greh->flags = tnl_flags_to_gre_flags(t->parms.o_flags);
561 greh->protocol = htons(type);
1294 562
1295 /* 563 memcpy(iph, &t->parms.iph, sizeof(struct iphdr));
1296 * Set the source hardware address.
1297 */
1298 564
565 /* Set the source hardware address. */
1299 if (saddr) 566 if (saddr)
1300 memcpy(&iph->saddr, saddr, 4); 567 memcpy(&iph->saddr, saddr, 4);
1301 if (daddr) 568 if (daddr)
@@ -1303,7 +570,7 @@ static int ipgre_header(struct sk_buff *skb, struct net_device *dev,
1303 if (iph->daddr) 570 if (iph->daddr)
1304 return t->hlen; 571 return t->hlen;
1305 572
1306 return -t->hlen; 573 return -(t->hlen + sizeof(*iph));
1307} 574}
1308 575
1309static int ipgre_header_parse(const struct sk_buff *skb, unsigned char *haddr) 576static int ipgre_header_parse(const struct sk_buff *skb, unsigned char *haddr)
@@ -1357,31 +624,21 @@ static int ipgre_close(struct net_device *dev)
1357 } 624 }
1358 return 0; 625 return 0;
1359} 626}
1360
1361#endif 627#endif
1362 628
1363static const struct net_device_ops ipgre_netdev_ops = { 629static const struct net_device_ops ipgre_netdev_ops = {
1364 .ndo_init = ipgre_tunnel_init, 630 .ndo_init = ipgre_tunnel_init,
1365 .ndo_uninit = ipgre_tunnel_uninit, 631 .ndo_uninit = ip_tunnel_uninit,
1366#ifdef CONFIG_NET_IPGRE_BROADCAST 632#ifdef CONFIG_NET_IPGRE_BROADCAST
1367 .ndo_open = ipgre_open, 633 .ndo_open = ipgre_open,
1368 .ndo_stop = ipgre_close, 634 .ndo_stop = ipgre_close,
1369#endif 635#endif
1370 .ndo_start_xmit = ipgre_tunnel_xmit, 636 .ndo_start_xmit = ipgre_xmit,
1371 .ndo_do_ioctl = ipgre_tunnel_ioctl, 637 .ndo_do_ioctl = ipgre_tunnel_ioctl,
1372 .ndo_change_mtu = ipgre_tunnel_change_mtu, 638 .ndo_change_mtu = ip_tunnel_change_mtu,
1373 .ndo_get_stats64 = ipgre_get_stats64, 639 .ndo_get_stats64 = ip_tunnel_get_stats64,
1374}; 640};
1375 641
1376static void ipgre_dev_free(struct net_device *dev)
1377{
1378 struct ip_tunnel *tunnel = netdev_priv(dev);
1379
1380 gro_cells_destroy(&tunnel->gro_cells);
1381 free_percpu(dev->tstats);
1382 free_netdev(dev);
1383}
1384
1385#define GRE_FEATURES (NETIF_F_SG | \ 642#define GRE_FEATURES (NETIF_F_SG | \
1386 NETIF_F_FRAGLIST | \ 643 NETIF_F_FRAGLIST | \
1387 NETIF_F_HIGHDMA | \ 644 NETIF_F_HIGHDMA | \
@@ -1390,35 +647,49 @@ static void ipgre_dev_free(struct net_device *dev)
1390static void ipgre_tunnel_setup(struct net_device *dev) 647static void ipgre_tunnel_setup(struct net_device *dev)
1391{ 648{
1392 dev->netdev_ops = &ipgre_netdev_ops; 649 dev->netdev_ops = &ipgre_netdev_ops;
1393 dev->destructor = ipgre_dev_free; 650 ip_tunnel_setup(dev, ipgre_net_id);
651}
1394 652
1395 dev->type = ARPHRD_IPGRE; 653static void __gre_tunnel_init(struct net_device *dev)
1396 dev->needed_headroom = LL_MAX_HEADER + sizeof(struct iphdr) + 4; 654{
655 struct ip_tunnel *tunnel;
656
657 tunnel = netdev_priv(dev);
658 tunnel->hlen = ip_gre_calc_hlen(tunnel->parms.o_flags);
659 tunnel->parms.iph.protocol = IPPROTO_GRE;
660
661 dev->needed_headroom = LL_MAX_HEADER + sizeof(struct iphdr) + 4;
1397 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr) - 4; 662 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr) - 4;
1398 dev->flags = IFF_NOARP;
1399 dev->iflink = 0; 663 dev->iflink = 0;
1400 dev->addr_len = 4;
1401 dev->features |= NETIF_F_NETNS_LOCAL;
1402 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1403 664
1404 dev->features |= GRE_FEATURES; 665 dev->features |= NETIF_F_NETNS_LOCAL | GRE_FEATURES;
1405 dev->hw_features |= GRE_FEATURES; 666 dev->hw_features |= GRE_FEATURES;
667
668 if (!(tunnel->parms.o_flags & TUNNEL_SEQ)) {
669 /* TCP offload with GRE SEQ is not supported. */
670 dev->features |= NETIF_F_GSO_SOFTWARE;
671 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
672 /* Can use a lockless transmit, unless we generate
673 * output sequences
674 */
675 dev->features |= NETIF_F_LLTX;
676 }
1406} 677}
1407 678
1408static int ipgre_tunnel_init(struct net_device *dev) 679static int ipgre_tunnel_init(struct net_device *dev)
1409{ 680{
1410 struct ip_tunnel *tunnel; 681 struct ip_tunnel *tunnel = netdev_priv(dev);
1411 struct iphdr *iph; 682 struct iphdr *iph = &tunnel->parms.iph;
1412 int err;
1413 683
1414 tunnel = netdev_priv(dev); 684 __gre_tunnel_init(dev);
1415 iph = &tunnel->parms.iph;
1416 685
1417 tunnel->dev = dev; 686 memcpy(dev->dev_addr, &iph->saddr, 4);
1418 strcpy(tunnel->parms.name, dev->name); 687 memcpy(dev->broadcast, &iph->daddr, 4);
1419 688
1420 memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4); 689 dev->type = ARPHRD_IPGRE;
1421 memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4); 690 dev->flags = IFF_NOARP;
691 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
692 dev->addr_len = 4;
1422 693
1423 if (iph->daddr) { 694 if (iph->daddr) {
1424#ifdef CONFIG_NET_IPGRE_BROADCAST 695#ifdef CONFIG_NET_IPGRE_BROADCAST
@@ -1432,106 +703,30 @@ static int ipgre_tunnel_init(struct net_device *dev)
1432 } else 703 } else
1433 dev->header_ops = &ipgre_header_ops; 704 dev->header_ops = &ipgre_header_ops;
1434 705
1435 dev->tstats = alloc_percpu(struct pcpu_tstats); 706 return ip_tunnel_init(dev);
1436 if (!dev->tstats)
1437 return -ENOMEM;
1438
1439 err = gro_cells_init(&tunnel->gro_cells, dev);
1440 if (err) {
1441 free_percpu(dev->tstats);
1442 return err;
1443 }
1444
1445 return 0;
1446} 707}
1447 708
1448static void ipgre_fb_tunnel_init(struct net_device *dev)
1449{
1450 struct ip_tunnel *tunnel = netdev_priv(dev);
1451 struct iphdr *iph = &tunnel->parms.iph;
1452
1453 tunnel->dev = dev;
1454 strcpy(tunnel->parms.name, dev->name);
1455
1456 iph->version = 4;
1457 iph->protocol = IPPROTO_GRE;
1458 iph->ihl = 5;
1459 tunnel->hlen = sizeof(struct iphdr) + 4;
1460
1461 dev_hold(dev);
1462}
1463
1464
1465static const struct gre_protocol ipgre_protocol = { 709static const struct gre_protocol ipgre_protocol = {
1466 .handler = ipgre_rcv, 710 .handler = ipgre_rcv,
1467 .err_handler = ipgre_err, 711 .err_handler = ipgre_err,
1468}; 712};
1469 713
1470static void ipgre_destroy_tunnels(struct ipgre_net *ign, struct list_head *head)
1471{
1472 int prio;
1473
1474 for (prio = 0; prio < 4; prio++) {
1475 int h;
1476 for (h = 0; h < HASH_SIZE; h++) {
1477 struct ip_tunnel *t;
1478
1479 t = rtnl_dereference(ign->tunnels[prio][h]);
1480
1481 while (t != NULL) {
1482 unregister_netdevice_queue(t->dev, head);
1483 t = rtnl_dereference(t->next);
1484 }
1485 }
1486 }
1487}
1488
1489static int __net_init ipgre_init_net(struct net *net) 714static int __net_init ipgre_init_net(struct net *net)
1490{ 715{
1491 struct ipgre_net *ign = net_generic(net, ipgre_net_id); 716 return ip_tunnel_init_net(net, ipgre_net_id, &ipgre_link_ops, NULL);
1492 int err;
1493
1494 ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "gre0",
1495 ipgre_tunnel_setup);
1496 if (!ign->fb_tunnel_dev) {
1497 err = -ENOMEM;
1498 goto err_alloc_dev;
1499 }
1500 dev_net_set(ign->fb_tunnel_dev, net);
1501
1502 ipgre_fb_tunnel_init(ign->fb_tunnel_dev);
1503 ign->fb_tunnel_dev->rtnl_link_ops = &ipgre_link_ops;
1504
1505 if ((err = register_netdev(ign->fb_tunnel_dev)))
1506 goto err_reg_dev;
1507
1508 rcu_assign_pointer(ign->tunnels_wc[0],
1509 netdev_priv(ign->fb_tunnel_dev));
1510 return 0;
1511
1512err_reg_dev:
1513 ipgre_dev_free(ign->fb_tunnel_dev);
1514err_alloc_dev:
1515 return err;
1516} 717}
1517 718
1518static void __net_exit ipgre_exit_net(struct net *net) 719static void __net_exit ipgre_exit_net(struct net *net)
1519{ 720{
1520 struct ipgre_net *ign; 721 struct ip_tunnel_net *itn = net_generic(net, ipgre_net_id);
1521 LIST_HEAD(list); 722 ip_tunnel_delete_net(itn);
1522
1523 ign = net_generic(net, ipgre_net_id);
1524 rtnl_lock();
1525 ipgre_destroy_tunnels(ign, &list);
1526 unregister_netdevice_many(&list);
1527 rtnl_unlock();
1528} 723}
1529 724
1530static struct pernet_operations ipgre_net_ops = { 725static struct pernet_operations ipgre_net_ops = {
1531 .init = ipgre_init_net, 726 .init = ipgre_init_net,
1532 .exit = ipgre_exit_net, 727 .exit = ipgre_exit_net,
1533 .id = &ipgre_net_id, 728 .id = &ipgre_net_id,
1534 .size = sizeof(struct ipgre_net), 729 .size = sizeof(struct ip_tunnel_net),
1535}; 730};
1536 731
1537static int ipgre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[]) 732static int ipgre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[])
@@ -1576,8 +771,8 @@ out:
1576 return ipgre_tunnel_validate(tb, data); 771 return ipgre_tunnel_validate(tb, data);
1577} 772}
1578 773
1579static void ipgre_netlink_parms(struct nlattr *data[], 774static void ipgre_netlink_parms(struct nlattr *data[], struct nlattr *tb[],
1580 struct ip_tunnel_parm *parms) 775 struct ip_tunnel_parm *parms)
1581{ 776{
1582 memset(parms, 0, sizeof(*parms)); 777 memset(parms, 0, sizeof(*parms));
1583 778
@@ -1590,10 +785,10 @@ static void ipgre_netlink_parms(struct nlattr *data[],
1590 parms->link = nla_get_u32(data[IFLA_GRE_LINK]); 785 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1591 786
1592 if (data[IFLA_GRE_IFLAGS]) 787 if (data[IFLA_GRE_IFLAGS])
1593 parms->i_flags = nla_get_be16(data[IFLA_GRE_IFLAGS]); 788 parms->i_flags = gre_flags_to_tnl_flags(nla_get_be16(data[IFLA_GRE_IFLAGS]));
1594 789
1595 if (data[IFLA_GRE_OFLAGS]) 790 if (data[IFLA_GRE_OFLAGS])
1596 parms->o_flags = nla_get_be16(data[IFLA_GRE_OFLAGS]); 791 parms->o_flags = gre_flags_to_tnl_flags(nla_get_be16(data[IFLA_GRE_OFLAGS]));
1597 792
1598 if (data[IFLA_GRE_IKEY]) 793 if (data[IFLA_GRE_IKEY])
1599 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]); 794 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
@@ -1617,148 +812,46 @@ static void ipgre_netlink_parms(struct nlattr *data[],
1617 parms->iph.frag_off = htons(IP_DF); 812 parms->iph.frag_off = htons(IP_DF);
1618} 813}
1619 814
1620static int ipgre_tap_init(struct net_device *dev) 815static int gre_tap_init(struct net_device *dev)
1621{ 816{
1622 struct ip_tunnel *tunnel; 817 __gre_tunnel_init(dev);
1623
1624 tunnel = netdev_priv(dev);
1625
1626 tunnel->dev = dev;
1627 strcpy(tunnel->parms.name, dev->name);
1628
1629 ipgre_tunnel_bind_dev(dev);
1630 818
1631 dev->tstats = alloc_percpu(struct pcpu_tstats); 819 return ip_tunnel_init(dev);
1632 if (!dev->tstats)
1633 return -ENOMEM;
1634
1635 return 0;
1636} 820}
1637 821
1638static const struct net_device_ops ipgre_tap_netdev_ops = { 822static const struct net_device_ops gre_tap_netdev_ops = {
1639 .ndo_init = ipgre_tap_init, 823 .ndo_init = gre_tap_init,
1640 .ndo_uninit = ipgre_tunnel_uninit, 824 .ndo_uninit = ip_tunnel_uninit,
1641 .ndo_start_xmit = ipgre_tunnel_xmit, 825 .ndo_start_xmit = gre_tap_xmit,
1642 .ndo_set_mac_address = eth_mac_addr, 826 .ndo_set_mac_address = eth_mac_addr,
1643 .ndo_validate_addr = eth_validate_addr, 827 .ndo_validate_addr = eth_validate_addr,
1644 .ndo_change_mtu = ipgre_tunnel_change_mtu, 828 .ndo_change_mtu = ip_tunnel_change_mtu,
1645 .ndo_get_stats64 = ipgre_get_stats64, 829 .ndo_get_stats64 = ip_tunnel_get_stats64,
1646}; 830};
1647 831
1648static void ipgre_tap_setup(struct net_device *dev) 832static void ipgre_tap_setup(struct net_device *dev)
1649{ 833{
1650
1651 ether_setup(dev); 834 ether_setup(dev);
1652 835 dev->netdev_ops = &gre_tap_netdev_ops;
1653 dev->netdev_ops = &ipgre_tap_netdev_ops; 836 ip_tunnel_setup(dev, gre_tap_net_id);
1654 dev->destructor = ipgre_dev_free;
1655
1656 dev->iflink = 0;
1657 dev->features |= NETIF_F_NETNS_LOCAL;
1658
1659 dev->features |= GRE_FEATURES;
1660 dev->hw_features |= GRE_FEATURES;
1661} 837}
1662 838
1663static int ipgre_newlink(struct net *src_net, struct net_device *dev, struct nlattr *tb[], 839static int ipgre_newlink(struct net *src_net, struct net_device *dev,
1664 struct nlattr *data[]) 840 struct nlattr *tb[], struct nlattr *data[])
1665{ 841{
1666 struct ip_tunnel *nt; 842 struct ip_tunnel_parm p;
1667 struct net *net = dev_net(dev);
1668 struct ipgre_net *ign = net_generic(net, ipgre_net_id);
1669 int mtu;
1670 int err;
1671
1672 nt = netdev_priv(dev);
1673 ipgre_netlink_parms(data, &nt->parms);
1674
1675 if (ipgre_tunnel_find(net, &nt->parms, dev->type))
1676 return -EEXIST;
1677
1678 if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1679 eth_hw_addr_random(dev);
1680
1681 mtu = ipgre_tunnel_bind_dev(dev);
1682 if (!tb[IFLA_MTU])
1683 dev->mtu = mtu;
1684
1685 /* Can use a lockless transmit, unless we generate output sequences */
1686 if (!(nt->parms.o_flags & GRE_SEQ))
1687 dev->features |= NETIF_F_LLTX;
1688
1689 err = register_netdevice(dev);
1690 if (err)
1691 goto out;
1692
1693 dev_hold(dev);
1694 ipgre_tunnel_link(ign, nt);
1695 843
1696out: 844 ipgre_netlink_parms(data, tb, &p);
1697 return err; 845 return ip_tunnel_newlink(dev, tb, &p);
1698} 846}
1699 847
1700static int ipgre_changelink(struct net_device *dev, struct nlattr *tb[], 848static int ipgre_changelink(struct net_device *dev, struct nlattr *tb[],
1701 struct nlattr *data[]) 849 struct nlattr *data[])
1702{ 850{
1703 struct ip_tunnel *t, *nt;
1704 struct net *net = dev_net(dev);
1705 struct ipgre_net *ign = net_generic(net, ipgre_net_id);
1706 struct ip_tunnel_parm p; 851 struct ip_tunnel_parm p;
1707 int mtu;
1708
1709 if (dev == ign->fb_tunnel_dev)
1710 return -EINVAL;
1711
1712 nt = netdev_priv(dev);
1713 ipgre_netlink_parms(data, &p);
1714
1715 t = ipgre_tunnel_locate(net, &p, 0);
1716
1717 if (t) {
1718 if (t->dev != dev)
1719 return -EEXIST;
1720 } else {
1721 t = nt;
1722
1723 if (dev->type != ARPHRD_ETHER) {
1724 unsigned int nflags = 0;
1725
1726 if (ipv4_is_multicast(p.iph.daddr))
1727 nflags = IFF_BROADCAST;
1728 else if (p.iph.daddr)
1729 nflags = IFF_POINTOPOINT;
1730
1731 if ((dev->flags ^ nflags) &
1732 (IFF_POINTOPOINT | IFF_BROADCAST))
1733 return -EINVAL;
1734 }
1735 852
1736 ipgre_tunnel_unlink(ign, t); 853 ipgre_netlink_parms(data, tb, &p);
1737 t->parms.iph.saddr = p.iph.saddr; 854 return ip_tunnel_changelink(dev, tb, &p);
1738 t->parms.iph.daddr = p.iph.daddr;
1739 t->parms.i_key = p.i_key;
1740 if (dev->type != ARPHRD_ETHER) {
1741 memcpy(dev->dev_addr, &p.iph.saddr, 4);
1742 memcpy(dev->broadcast, &p.iph.daddr, 4);
1743 }
1744 ipgre_tunnel_link(ign, t);
1745 netdev_state_change(dev);
1746 }
1747
1748 t->parms.o_key = p.o_key;
1749 t->parms.iph.ttl = p.iph.ttl;
1750 t->parms.iph.tos = p.iph.tos;
1751 t->parms.iph.frag_off = p.iph.frag_off;
1752
1753 if (t->parms.link != p.link) {
1754 t->parms.link = p.link;
1755 mtu = ipgre_tunnel_bind_dev(dev);
1756 if (!tb[IFLA_MTU])
1757 dev->mtu = mtu;
1758 netdev_state_change(dev);
1759 }
1760
1761 return 0;
1762} 855}
1763 856
1764static size_t ipgre_get_size(const struct net_device *dev) 857static size_t ipgre_get_size(const struct net_device *dev)
@@ -1793,8 +886,8 @@ static int ipgre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1793 struct ip_tunnel_parm *p = &t->parms; 886 struct ip_tunnel_parm *p = &t->parms;
1794 887
1795 if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) || 888 if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1796 nla_put_be16(skb, IFLA_GRE_IFLAGS, p->i_flags) || 889 nla_put_be16(skb, IFLA_GRE_IFLAGS, tnl_flags_to_gre_flags(p->i_flags)) ||
1797 nla_put_be16(skb, IFLA_GRE_OFLAGS, p->o_flags) || 890 nla_put_be16(skb, IFLA_GRE_OFLAGS, tnl_flags_to_gre_flags(p->o_flags)) ||
1798 nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) || 891 nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1799 nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) || 892 nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
1800 nla_put_be32(skb, IFLA_GRE_LOCAL, p->iph.saddr) || 893 nla_put_be32(skb, IFLA_GRE_LOCAL, p->iph.saddr) ||
@@ -1832,6 +925,7 @@ static struct rtnl_link_ops ipgre_link_ops __read_mostly = {
1832 .validate = ipgre_tunnel_validate, 925 .validate = ipgre_tunnel_validate,
1833 .newlink = ipgre_newlink, 926 .newlink = ipgre_newlink,
1834 .changelink = ipgre_changelink, 927 .changelink = ipgre_changelink,
928 .dellink = ip_tunnel_dellink,
1835 .get_size = ipgre_get_size, 929 .get_size = ipgre_get_size,
1836 .fill_info = ipgre_fill_info, 930 .fill_info = ipgre_fill_info,
1837}; 931};
@@ -1845,13 +939,28 @@ static struct rtnl_link_ops ipgre_tap_ops __read_mostly = {
1845 .validate = ipgre_tap_validate, 939 .validate = ipgre_tap_validate,
1846 .newlink = ipgre_newlink, 940 .newlink = ipgre_newlink,
1847 .changelink = ipgre_changelink, 941 .changelink = ipgre_changelink,
942 .dellink = ip_tunnel_dellink,
1848 .get_size = ipgre_get_size, 943 .get_size = ipgre_get_size,
1849 .fill_info = ipgre_fill_info, 944 .fill_info = ipgre_fill_info,
1850}; 945};
1851 946
1852/* 947static int __net_init ipgre_tap_init_net(struct net *net)
1853 * And now the modules code and kernel interface. 948{
1854 */ 949 return ip_tunnel_init_net(net, gre_tap_net_id, &ipgre_tap_ops, NULL);
950}
951
952static void __net_exit ipgre_tap_exit_net(struct net *net)
953{
954 struct ip_tunnel_net *itn = net_generic(net, gre_tap_net_id);
955 ip_tunnel_delete_net(itn);
956}
957
958static struct pernet_operations ipgre_tap_net_ops = {
959 .init = ipgre_tap_init_net,
960 .exit = ipgre_tap_exit_net,
961 .id = &gre_tap_net_id,
962 .size = sizeof(struct ip_tunnel_net),
963};
1855 964
1856static int __init ipgre_init(void) 965static int __init ipgre_init(void)
1857{ 966{
@@ -1863,6 +972,10 @@ static int __init ipgre_init(void)
1863 if (err < 0) 972 if (err < 0)
1864 return err; 973 return err;
1865 974
975 err = register_pernet_device(&ipgre_tap_net_ops);
976 if (err < 0)
977 goto pnet_tap_faied;
978
1866 err = gre_add_protocol(&ipgre_protocol, GREPROTO_CISCO); 979 err = gre_add_protocol(&ipgre_protocol, GREPROTO_CISCO);
1867 if (err < 0) { 980 if (err < 0) {
1868 pr_info("%s: can't add protocol\n", __func__); 981 pr_info("%s: can't add protocol\n", __func__);
@@ -1877,16 +990,17 @@ static int __init ipgre_init(void)
1877 if (err < 0) 990 if (err < 0)
1878 goto tap_ops_failed; 991 goto tap_ops_failed;
1879 992
1880out: 993 return 0;
1881 return err;
1882 994
1883tap_ops_failed: 995tap_ops_failed:
1884 rtnl_link_unregister(&ipgre_link_ops); 996 rtnl_link_unregister(&ipgre_link_ops);
1885rtnl_link_failed: 997rtnl_link_failed:
1886 gre_del_protocol(&ipgre_protocol, GREPROTO_CISCO); 998 gre_del_protocol(&ipgre_protocol, GREPROTO_CISCO);
1887add_proto_failed: 999add_proto_failed:
1000 unregister_pernet_device(&ipgre_tap_net_ops);
1001pnet_tap_faied:
1888 unregister_pernet_device(&ipgre_net_ops); 1002 unregister_pernet_device(&ipgre_net_ops);
1889 goto out; 1003 return err;
1890} 1004}
1891 1005
1892static void __exit ipgre_fini(void) 1006static void __exit ipgre_fini(void)
@@ -1895,6 +1009,7 @@ static void __exit ipgre_fini(void)
1895 rtnl_link_unregister(&ipgre_link_ops); 1009 rtnl_link_unregister(&ipgre_link_ops);
1896 if (gre_del_protocol(&ipgre_protocol, GREPROTO_CISCO) < 0) 1010 if (gre_del_protocol(&ipgre_protocol, GREPROTO_CISCO) < 0)
1897 pr_info("%s: can't remove protocol\n", __func__); 1011 pr_info("%s: can't remove protocol\n", __func__);
1012 unregister_pernet_device(&ipgre_tap_net_ops);
1898 unregister_pernet_device(&ipgre_net_ops); 1013 unregister_pernet_device(&ipgre_net_ops);
1899} 1014}
1900 1015
@@ -1904,3 +1019,4 @@ MODULE_LICENSE("GPL");
1904MODULE_ALIAS_RTNL_LINK("gre"); 1019MODULE_ALIAS_RTNL_LINK("gre");
1905MODULE_ALIAS_RTNL_LINK("gretap"); 1020MODULE_ALIAS_RTNL_LINK("gretap");
1906MODULE_ALIAS_NETDEV("gre0"); 1021MODULE_ALIAS_NETDEV("gre0");
1022MODULE_ALIAS_NETDEV("gretap0");