Commit | Line | Data |
---|---|---|
1181412c S |
1 | /* |
2 | * Linux NET3: IP/IP protocol decoder modified to support | |
3 | * virtual tunnel interface | |
4 | * | |
5 | * Authors: | |
6 | * Saurabh Mohan (saurabh.mohan@vyatta.com) 05/07/2012 | |
7 | * | |
8 | * This program is free software; you can redistribute it and/or | |
9 | * modify it under the terms of the GNU General Public License | |
10 | * as published by the Free Software Foundation; either version | |
11 | * 2 of the License, or (at your option) any later version. | |
12 | * | |
13 | */ | |
14 | ||
15 | /* | |
16 | This version of net/ipv4/ip_vti.c is cloned of net/ipv4/ipip.c | |
17 | ||
18 | For comments look at net/ipv4/ip_gre.c --ANK | |
19 | */ | |
20 | ||
21 | ||
22 | #include <linux/capability.h> | |
23 | #include <linux/module.h> | |
24 | #include <linux/types.h> | |
25 | #include <linux/kernel.h> | |
26 | #include <linux/uaccess.h> | |
27 | #include <linux/skbuff.h> | |
28 | #include <linux/netdevice.h> | |
29 | #include <linux/in.h> | |
30 | #include <linux/tcp.h> | |
31 | #include <linux/udp.h> | |
32 | #include <linux/if_arp.h> | |
33 | #include <linux/mroute.h> | |
34 | #include <linux/init.h> | |
35 | #include <linux/netfilter_ipv4.h> | |
36 | #include <linux/if_ether.h> | |
37 | ||
38 | #include <net/sock.h> | |
39 | #include <net/ip.h> | |
40 | #include <net/icmp.h> | |
c5441932 | 41 | #include <net/ip_tunnels.h> |
1181412c S |
42 | #include <net/inet_ecn.h> |
43 | #include <net/xfrm.h> | |
44 | #include <net/net_namespace.h> | |
45 | #include <net/netns/generic.h> | |
46 | ||
1181412c S |
47 | static struct rtnl_link_ops vti_link_ops __read_mostly; |
48 | ||
49 | static int vti_net_id __read_mostly; | |
1181412c | 50 | static int vti_tunnel_init(struct net_device *dev); |
1181412c S |
51 | |
52 | static int vti_err(struct sk_buff *skb, u32 info) | |
53 | { | |
54 | ||
55 | /* All the routers (except for Linux) return only | |
56 | * 8 bytes of packet payload. It means, that precise relaying of | |
57 | * ICMP in the real Internet is absolutely infeasible. | |
58 | */ | |
b9959fd3 AW |
59 | struct net *net = dev_net(skb->dev); |
60 | struct ip_tunnel_net *itn = net_generic(net, vti_net_id); | |
1181412c S |
61 | struct iphdr *iph = (struct iphdr *)skb->data; |
62 | const int type = icmp_hdr(skb)->type; | |
63 | const int code = icmp_hdr(skb)->code; | |
64 | struct ip_tunnel *t; | |
65 | int err; | |
66 | ||
67 | switch (type) { | |
68 | default: | |
69 | case ICMP_PARAMETERPROB: | |
70 | return 0; | |
71 | ||
72 | case ICMP_DEST_UNREACH: | |
73 | switch (code) { | |
74 | case ICMP_SR_FAILED: | |
75 | case ICMP_PORT_UNREACH: | |
76 | /* Impossible event. */ | |
77 | return 0; | |
78 | default: | |
79 | /* All others are translated to HOST_UNREACH. */ | |
80 | break; | |
81 | } | |
82 | break; | |
83 | case ICMP_TIME_EXCEEDED: | |
84 | if (code != ICMP_EXC_TTL) | |
85 | return 0; | |
86 | break; | |
87 | } | |
88 | ||
89 | err = -ENOENT; | |
90 | ||
b9959fd3 AW |
91 | t = ip_tunnel_lookup(itn, skb->dev->ifindex, TUNNEL_NO_KEY, |
92 | iph->daddr, iph->saddr, 0); | |
1181412c S |
93 | if (t == NULL) |
94 | goto out; | |
95 | ||
96 | if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) { | |
97 | ipv4_update_pmtu(skb, dev_net(skb->dev), info, | |
98 | t->parms.link, 0, IPPROTO_IPIP, 0); | |
99 | err = 0; | |
100 | goto out; | |
101 | } | |
102 | ||
103 | err = 0; | |
104 | if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED) | |
105 | goto out; | |
106 | ||
107 | if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO)) | |
108 | t->err_count++; | |
109 | else | |
110 | t->err_count = 1; | |
111 | t->err_time = jiffies; | |
112 | out: | |
1181412c S |
113 | return err; |
114 | } | |
115 | ||
116 | /* We dont digest the packet therefore let the packet pass */ | |
117 | static int vti_rcv(struct sk_buff *skb) | |
118 | { | |
119 | struct ip_tunnel *tunnel; | |
120 | const struct iphdr *iph = ip_hdr(skb); | |
b9959fd3 AW |
121 | struct net *net = dev_net(skb->dev); |
122 | struct ip_tunnel_net *itn = net_generic(net, vti_net_id); | |
1181412c | 123 | |
b9959fd3 AW |
124 | tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, TUNNEL_NO_KEY, |
125 | iph->saddr, iph->daddr, 0); | |
1181412c S |
126 | if (tunnel != NULL) { |
127 | struct pcpu_tstats *tstats; | |
128 | ||
b2942004 SM |
129 | if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) |
130 | return -1; | |
131 | ||
1181412c S |
132 | tstats = this_cpu_ptr(tunnel->dev->tstats); |
133 | u64_stats_update_begin(&tstats->syncp); | |
134 | tstats->rx_packets++; | |
135 | tstats->rx_bytes += skb->len; | |
136 | u64_stats_update_end(&tstats->syncp); | |
137 | ||
b2942004 SM |
138 | skb->mark = 0; |
139 | secpath_reset(skb); | |
1181412c | 140 | skb->dev = tunnel->dev; |
1181412c S |
141 | return 1; |
142 | } | |
1181412c S |
143 | |
144 | return -1; | |
145 | } | |
146 | ||
147 | /* This function assumes it is being called from dev_queue_xmit() | |
148 | * and that skb is filled properly by that function. | |
149 | */ | |
150 | ||
151 | static netdev_tx_t vti_tunnel_xmit(struct sk_buff *skb, struct net_device *dev) | |
152 | { | |
153 | struct ip_tunnel *tunnel = netdev_priv(dev); | |
1181412c S |
154 | struct iphdr *tiph = &tunnel->parms.iph; |
155 | u8 tos; | |
156 | struct rtable *rt; /* Route to the other host */ | |
157 | struct net_device *tdev; /* Device to other host */ | |
158 | struct iphdr *old_iph = ip_hdr(skb); | |
159 | __be32 dst = tiph->daddr; | |
160 | struct flowi4 fl4; | |
b9959fd3 | 161 | int err; |
1181412c S |
162 | |
163 | if (skb->protocol != htons(ETH_P_IP)) | |
164 | goto tx_error; | |
165 | ||
166 | tos = old_iph->tos; | |
167 | ||
168 | memset(&fl4, 0, sizeof(fl4)); | |
169 | flowi4_init_output(&fl4, tunnel->parms.link, | |
8437e761 | 170 | be32_to_cpu(tunnel->parms.i_key), RT_TOS(tos), |
1181412c S |
171 | RT_SCOPE_UNIVERSE, |
172 | IPPROTO_IPIP, 0, | |
173 | dst, tiph->saddr, 0, 0); | |
174 | rt = ip_route_output_key(dev_net(dev), &fl4); | |
175 | if (IS_ERR(rt)) { | |
176 | dev->stats.tx_carrier_errors++; | |
177 | goto tx_error_icmp; | |
178 | } | |
179 | /* if there is no transform then this tunnel is not functional. | |
180 | * Or if the xfrm is not mode tunnel. | |
181 | */ | |
182 | if (!rt->dst.xfrm || | |
183 | rt->dst.xfrm->props.mode != XFRM_MODE_TUNNEL) { | |
184 | dev->stats.tx_carrier_errors++; | |
185 | goto tx_error_icmp; | |
186 | } | |
187 | tdev = rt->dst.dev; | |
188 | ||
189 | if (tdev == dev) { | |
190 | ip_rt_put(rt); | |
191 | dev->stats.collisions++; | |
192 | goto tx_error; | |
193 | } | |
194 | ||
195 | if (tunnel->err_count > 0) { | |
196 | if (time_before(jiffies, | |
197 | tunnel->err_time + IPTUNNEL_ERR_TIMEO)) { | |
198 | tunnel->err_count--; | |
199 | dst_link_failure(skb); | |
200 | } else | |
201 | tunnel->err_count = 0; | |
202 | } | |
203 | ||
baafc77b | 204 | memset(IPCB(skb), 0, sizeof(*IPCB(skb))); |
1181412c S |
205 | skb_dst_drop(skb); |
206 | skb_dst_set(skb, &rt->dst); | |
207 | nf_reset(skb); | |
208 | skb->dev = skb_dst(skb)->dev; | |
209 | ||
b9959fd3 AW |
210 | err = dst_output(skb); |
211 | if (net_xmit_eval(err) == 0) | |
212 | err = skb->len; | |
213 | iptunnel_xmit_stats(err, &dev->stats, dev->tstats); | |
1181412c S |
214 | return NETDEV_TX_OK; |
215 | ||
216 | tx_error_icmp: | |
217 | dst_link_failure(skb); | |
218 | tx_error: | |
219 | dev->stats.tx_errors++; | |
220 | dev_kfree_skb(skb); | |
221 | return NETDEV_TX_OK; | |
222 | } | |
223 | ||
1181412c S |
224 | static int |
225 | vti_tunnel_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd) | |
226 | { | |
227 | int err = 0; | |
228 | struct ip_tunnel_parm p; | |
1181412c | 229 | |
b9959fd3 AW |
230 | if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) |
231 | return -EFAULT; | |
1181412c | 232 | |
b9959fd3 | 233 | if (cmd == SIOCADDTUNNEL || cmd == SIOCCHGTUNNEL) { |
1181412c S |
234 | if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPIP || |
235 | p.iph.ihl != 5) | |
b9959fd3 AW |
236 | return -EINVAL; |
237 | } | |
1181412c | 238 | |
b9959fd3 AW |
239 | err = ip_tunnel_ioctl(dev, &p, cmd); |
240 | if (err) | |
241 | return err; | |
1181412c | 242 | |
b9959fd3 AW |
243 | if (cmd != SIOCDELTUNNEL) { |
244 | p.i_flags |= GRE_KEY | VTI_ISVTI; | |
245 | p.o_flags |= GRE_KEY; | |
1181412c S |
246 | } |
247 | ||
b9959fd3 AW |
248 | if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) |
249 | return -EFAULT; | |
1181412c S |
250 | return 0; |
251 | } | |
252 | ||
253 | static const struct net_device_ops vti_netdev_ops = { | |
254 | .ndo_init = vti_tunnel_init, | |
b9959fd3 | 255 | .ndo_uninit = ip_tunnel_uninit, |
1181412c S |
256 | .ndo_start_xmit = vti_tunnel_xmit, |
257 | .ndo_do_ioctl = vti_tunnel_ioctl, | |
b9959fd3 | 258 | .ndo_change_mtu = ip_tunnel_change_mtu, |
f61dd388 | 259 | .ndo_get_stats64 = ip_tunnel_get_stats64, |
1181412c S |
260 | }; |
261 | ||
b9959fd3 | 262 | static void vti_tunnel_setup(struct net_device *dev) |
1181412c | 263 | { |
b9959fd3 AW |
264 | dev->netdev_ops = &vti_netdev_ops; |
265 | ip_tunnel_setup(dev, vti_net_id); | |
1181412c S |
266 | } |
267 | ||
b9959fd3 | 268 | static int vti_tunnel_init(struct net_device *dev) |
1181412c | 269 | { |
b9959fd3 AW |
270 | struct ip_tunnel *tunnel = netdev_priv(dev); |
271 | struct iphdr *iph = &tunnel->parms.iph; | |
272 | ||
273 | memcpy(dev->dev_addr, &iph->saddr, 4); | |
274 | memcpy(dev->broadcast, &iph->daddr, 4); | |
1181412c S |
275 | |
276 | dev->type = ARPHRD_TUNNEL; | |
277 | dev->hard_header_len = LL_MAX_HEADER + sizeof(struct iphdr); | |
278 | dev->mtu = ETH_DATA_LEN; | |
279 | dev->flags = IFF_NOARP; | |
280 | dev->iflink = 0; | |
281 | dev->addr_len = 4; | |
282 | dev->features |= NETIF_F_NETNS_LOCAL; | |
283 | dev->features |= NETIF_F_LLTX; | |
284 | dev->priv_flags &= ~IFF_XMIT_DST_RELEASE; | |
1181412c | 285 | |
b9959fd3 | 286 | return ip_tunnel_init(dev); |
1181412c S |
287 | } |
288 | ||
b9959fd3 | 289 | static void __net_init vti_fb_tunnel_init(struct net_device *dev) |
1181412c S |
290 | { |
291 | struct ip_tunnel *tunnel = netdev_priv(dev); | |
292 | struct iphdr *iph = &tunnel->parms.iph; | |
1181412c | 293 | |
1181412c S |
294 | iph->version = 4; |
295 | iph->protocol = IPPROTO_IPIP; | |
296 | iph->ihl = 5; | |
1181412c S |
297 | } |
298 | ||
299 | static struct xfrm_tunnel vti_handler __read_mostly = { | |
300 | .handler = vti_rcv, | |
301 | .err_handler = vti_err, | |
302 | .priority = 1, | |
303 | }; | |
304 | ||
1181412c S |
305 | static int __net_init vti_init_net(struct net *net) |
306 | { | |
307 | int err; | |
b9959fd3 | 308 | struct ip_tunnel_net *itn; |
1181412c | 309 | |
b9959fd3 | 310 | err = ip_tunnel_init_net(net, vti_net_id, &vti_link_ops, "ip_vti0"); |
1181412c | 311 | if (err) |
b9959fd3 AW |
312 | return err; |
313 | itn = net_generic(net, vti_net_id); | |
314 | vti_fb_tunnel_init(itn->fb_tunnel_dev); | |
1181412c | 315 | return 0; |
1181412c S |
316 | } |
317 | ||
318 | static void __net_exit vti_exit_net(struct net *net) | |
319 | { | |
b9959fd3 AW |
320 | struct ip_tunnel_net *itn = net_generic(net, vti_net_id); |
321 | ip_tunnel_delete_net(itn); | |
1181412c S |
322 | } |
323 | ||
324 | static struct pernet_operations vti_net_ops = { | |
325 | .init = vti_init_net, | |
326 | .exit = vti_exit_net, | |
327 | .id = &vti_net_id, | |
b9959fd3 | 328 | .size = sizeof(struct ip_tunnel_net), |
1181412c S |
329 | }; |
330 | ||
331 | static int vti_tunnel_validate(struct nlattr *tb[], struct nlattr *data[]) | |
332 | { | |
333 | return 0; | |
334 | } | |
335 | ||
336 | static void vti_netlink_parms(struct nlattr *data[], | |
337 | struct ip_tunnel_parm *parms) | |
338 | { | |
339 | memset(parms, 0, sizeof(*parms)); | |
340 | ||
341 | parms->iph.protocol = IPPROTO_IPIP; | |
342 | ||
343 | if (!data) | |
344 | return; | |
345 | ||
346 | if (data[IFLA_VTI_LINK]) | |
347 | parms->link = nla_get_u32(data[IFLA_VTI_LINK]); | |
348 | ||
349 | if (data[IFLA_VTI_IKEY]) | |
350 | parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]); | |
351 | ||
352 | if (data[IFLA_VTI_OKEY]) | |
353 | parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]); | |
354 | ||
355 | if (data[IFLA_VTI_LOCAL]) | |
356 | parms->iph.saddr = nla_get_be32(data[IFLA_VTI_LOCAL]); | |
357 | ||
358 | if (data[IFLA_VTI_REMOTE]) | |
359 | parms->iph.daddr = nla_get_be32(data[IFLA_VTI_REMOTE]); | |
360 | ||
361 | } | |
362 | ||
363 | static int vti_newlink(struct net *src_net, struct net_device *dev, | |
364 | struct nlattr *tb[], struct nlattr *data[]) | |
365 | { | |
b9959fd3 | 366 | struct ip_tunnel_parm parms; |
1181412c | 367 | |
b9959fd3 AW |
368 | vti_netlink_parms(data, &parms); |
369 | return ip_tunnel_newlink(dev, tb, &parms); | |
1181412c S |
370 | } |
371 | ||
372 | static int vti_changelink(struct net_device *dev, struct nlattr *tb[], | |
373 | struct nlattr *data[]) | |
374 | { | |
1181412c | 375 | struct ip_tunnel_parm p; |
1181412c | 376 | |
1181412c | 377 | vti_netlink_parms(data, &p); |
b9959fd3 | 378 | return ip_tunnel_changelink(dev, tb, &p); |
1181412c S |
379 | } |
380 | ||
381 | static size_t vti_get_size(const struct net_device *dev) | |
382 | { | |
383 | return | |
384 | /* IFLA_VTI_LINK */ | |
385 | nla_total_size(4) + | |
386 | /* IFLA_VTI_IKEY */ | |
387 | nla_total_size(4) + | |
388 | /* IFLA_VTI_OKEY */ | |
389 | nla_total_size(4) + | |
390 | /* IFLA_VTI_LOCAL */ | |
391 | nla_total_size(4) + | |
392 | /* IFLA_VTI_REMOTE */ | |
393 | nla_total_size(4) + | |
394 | 0; | |
395 | } | |
396 | ||
397 | static int vti_fill_info(struct sk_buff *skb, const struct net_device *dev) | |
398 | { | |
399 | struct ip_tunnel *t = netdev_priv(dev); | |
400 | struct ip_tunnel_parm *p = &t->parms; | |
401 | ||
402 | nla_put_u32(skb, IFLA_VTI_LINK, p->link); | |
403 | nla_put_be32(skb, IFLA_VTI_IKEY, p->i_key); | |
404 | nla_put_be32(skb, IFLA_VTI_OKEY, p->o_key); | |
405 | nla_put_be32(skb, IFLA_VTI_LOCAL, p->iph.saddr); | |
406 | nla_put_be32(skb, IFLA_VTI_REMOTE, p->iph.daddr); | |
407 | ||
408 | return 0; | |
409 | } | |
410 | ||
411 | static const struct nla_policy vti_policy[IFLA_VTI_MAX + 1] = { | |
412 | [IFLA_VTI_LINK] = { .type = NLA_U32 }, | |
413 | [IFLA_VTI_IKEY] = { .type = NLA_U32 }, | |
414 | [IFLA_VTI_OKEY] = { .type = NLA_U32 }, | |
415 | [IFLA_VTI_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) }, | |
416 | [IFLA_VTI_REMOTE] = { .len = FIELD_SIZEOF(struct iphdr, daddr) }, | |
417 | }; | |
418 | ||
419 | static struct rtnl_link_ops vti_link_ops __read_mostly = { | |
420 | .kind = "vti", | |
421 | .maxtype = IFLA_VTI_MAX, | |
422 | .policy = vti_policy, | |
423 | .priv_size = sizeof(struct ip_tunnel), | |
424 | .setup = vti_tunnel_setup, | |
425 | .validate = vti_tunnel_validate, | |
426 | .newlink = vti_newlink, | |
427 | .changelink = vti_changelink, | |
428 | .get_size = vti_get_size, | |
429 | .fill_info = vti_fill_info, | |
430 | }; | |
431 | ||
432 | static int __init vti_init(void) | |
433 | { | |
434 | int err; | |
435 | ||
436 | pr_info("IPv4 over IPSec tunneling driver\n"); | |
437 | ||
438 | err = register_pernet_device(&vti_net_ops); | |
439 | if (err < 0) | |
440 | return err; | |
441 | err = xfrm4_mode_tunnel_input_register(&vti_handler); | |
442 | if (err < 0) { | |
443 | unregister_pernet_device(&vti_net_ops); | |
b9959fd3 | 444 | pr_info("vti init: can't register tunnel\n"); |
1181412c S |
445 | } |
446 | ||
447 | err = rtnl_link_register(&vti_link_ops); | |
448 | if (err < 0) | |
449 | goto rtnl_link_failed; | |
450 | ||
451 | return err; | |
452 | ||
453 | rtnl_link_failed: | |
454 | xfrm4_mode_tunnel_input_deregister(&vti_handler); | |
455 | unregister_pernet_device(&vti_net_ops); | |
456 | return err; | |
457 | } | |
458 | ||
459 | static void __exit vti_fini(void) | |
460 | { | |
461 | rtnl_link_unregister(&vti_link_ops); | |
462 | if (xfrm4_mode_tunnel_input_deregister(&vti_handler)) | |
463 | pr_info("vti close: can't deregister tunnel\n"); | |
464 | ||
465 | unregister_pernet_device(&vti_net_ops); | |
466 | } | |
467 | ||
468 | module_init(vti_init); | |
469 | module_exit(vti_fini); | |
470 | MODULE_LICENSE("GPL"); | |
471 | MODULE_ALIAS_RTNL_LINK("vti"); | |
472 | MODULE_ALIAS_NETDEV("ip_vti0"); |