Commit | Line | Data |
---|---|---|
1da177e4 | 1 | /* |
e905a9ed | 2 | * Linux NET3: GRE over IP protocol decoder. |
1da177e4 LT |
3 | * |
4 | * Authors: Alexey Kuznetsov (kuznet@ms2.inr.ac.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 | ||
afd46503 JP |
13 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt |
14 | ||
4fc268d2 | 15 | #include <linux/capability.h> |
1da177e4 LT |
16 | #include <linux/module.h> |
17 | #include <linux/types.h> | |
1da177e4 | 18 | #include <linux/kernel.h> |
5a0e3ad6 | 19 | #include <linux/slab.h> |
1da177e4 LT |
20 | #include <asm/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> | |
2e15ea39 | 27 | #include <linux/if_vlan.h> |
1da177e4 LT |
28 | #include <linux/init.h> |
29 | #include <linux/in6.h> | |
30 | #include <linux/inetdevice.h> | |
31 | #include <linux/igmp.h> | |
32 | #include <linux/netfilter_ipv4.h> | |
e1a80002 | 33 | #include <linux/etherdevice.h> |
46f25dff | 34 | #include <linux/if_ether.h> |
1da177e4 LT |
35 | |
36 | #include <net/sock.h> | |
37 | #include <net/ip.h> | |
38 | #include <net/icmp.h> | |
39 | #include <net/protocol.h> | |
c5441932 | 40 | #include <net/ip_tunnels.h> |
1da177e4 LT |
41 | #include <net/arp.h> |
42 | #include <net/checksum.h> | |
43 | #include <net/dsfield.h> | |
44 | #include <net/inet_ecn.h> | |
45 | #include <net/xfrm.h> | |
59a4c759 PE |
46 | #include <net/net_namespace.h> |
47 | #include <net/netns/generic.h> | |
c19e654d | 48 | #include <net/rtnetlink.h> |
00959ade | 49 | #include <net/gre.h> |
2e15ea39 | 50 | #include <net/dst_metadata.h> |
1da177e4 | 51 | |
dfd56b8b | 52 | #if IS_ENABLED(CONFIG_IPV6) |
1da177e4 LT |
53 | #include <net/ipv6.h> |
54 | #include <net/ip6_fib.h> | |
55 | #include <net/ip6_route.h> | |
56 | #endif | |
57 | ||
58 | /* | |
59 | Problems & solutions | |
60 | -------------------- | |
61 | ||
62 | 1. The most important issue is detecting local dead loops. | |
63 | They would cause complete host lockup in transmit, which | |
64 | would be "resolved" by stack overflow or, if queueing is enabled, | |
65 | with infinite looping in net_bh. | |
66 | ||
67 | We cannot track such dead loops during route installation, | |
68 | it is infeasible task. The most general solutions would be | |
69 | to keep skb->encapsulation counter (sort of local ttl), | |
6d0722a2 | 70 | and silently drop packet when it expires. It is a good |
bff52857 | 71 | solution, but it supposes maintaining new variable in ALL |
1da177e4 LT |
72 | skb, even if no tunneling is used. |
73 | ||
6d0722a2 ED |
74 | Current solution: xmit_recursion breaks dead loops. This is a percpu |
75 | counter, since when we enter the first ndo_xmit(), cpu migration is | |
76 | forbidden. We force an exit if this counter reaches RECURSION_LIMIT | |
1da177e4 LT |
77 | |
78 | 2. Networking dead loops would not kill routers, but would really | |
79 | kill network. IP hop limit plays role of "t->recursion" in this case, | |
80 | if we copy it from packet being encapsulated to upper header. | |
81 | It is very good solution, but it introduces two problems: | |
82 | ||
83 | - Routing protocols, using packets with ttl=1 (OSPF, RIP2), | |
84 | do not work over tunnels. | |
85 | - traceroute does not work. I planned to relay ICMP from tunnel, | |
86 | so that this problem would be solved and traceroute output | |
87 | would even more informative. This idea appeared to be wrong: | |
88 | only Linux complies to rfc1812 now (yes, guys, Linux is the only | |
89 | true router now :-)), all routers (at least, in neighbourhood of mine) | |
90 | return only 8 bytes of payload. It is the end. | |
91 | ||
92 | Hence, if we want that OSPF worked or traceroute said something reasonable, | |
93 | we should search for another solution. | |
94 | ||
95 | One of them is to parse packet trying to detect inner encapsulation | |
96 | made by our node. It is difficult or even impossible, especially, | |
bff52857 | 97 | taking into account fragmentation. TO be short, ttl is not solution at all. |
1da177e4 LT |
98 | |
99 | Current solution: The solution was UNEXPECTEDLY SIMPLE. | |
100 | We force DF flag on tunnels with preconfigured hop limit, | |
101 | that is ALL. :-) Well, it does not remove the problem completely, | |
102 | but exponential growth of network traffic is changed to linear | |
103 | (branches, that exceed pmtu are pruned) and tunnel mtu | |
bff52857 | 104 | rapidly degrades to value <68, where looping stops. |
1da177e4 LT |
105 | Yes, it is not good if there exists a router in the loop, |
106 | which does not force DF, even when encapsulating packets have DF set. | |
107 | But it is not our problem! Nobody could accuse us, we made | |
108 | all that we could make. Even if it is your gated who injected | |
109 | fatal route to network, even if it were you who configured | |
110 | fatal static route: you are innocent. :-) | |
111 | ||
1da177e4 LT |
112 | Alexey Kuznetsov. |
113 | */ | |
114 | ||
eccc1bb8 | 115 | static bool log_ecn_error = true; |
116 | module_param(log_ecn_error, bool, 0644); | |
117 | MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN"); | |
118 | ||
c19e654d | 119 | static struct rtnl_link_ops ipgre_link_ops __read_mostly; |
1da177e4 | 120 | static int ipgre_tunnel_init(struct net_device *dev); |
eb8ce741 | 121 | |
f99189b1 | 122 | static int ipgre_net_id __read_mostly; |
c5441932 | 123 | static int gre_tap_net_id __read_mostly; |
1da177e4 | 124 | |
9f57c67c PS |
125 | static void ipgre_err(struct sk_buff *skb, u32 info, |
126 | const struct tnl_ptk_info *tpi) | |
1da177e4 | 127 | { |
1da177e4 | 128 | |
c5441932 PS |
129 | /* All the routers (except for Linux) return only |
130 | 8 bytes of packet payload. It means, that precise relaying of | |
131 | ICMP in the real Internet is absolutely infeasible. | |
1da177e4 | 132 | |
c5441932 PS |
133 | Moreover, Cisco "wise men" put GRE key to the third word |
134 | in GRE header. It makes impossible maintaining even soft | |
135 | state for keyed GRE tunnels with enabled checksum. Tell | |
136 | them "thank you". | |
1da177e4 | 137 | |
c5441932 PS |
138 | Well, I wonder, rfc1812 was written by Cisco employee, |
139 | what the hell these idiots break standards established | |
140 | by themselves??? | |
141 | */ | |
142 | struct net *net = dev_net(skb->dev); | |
143 | struct ip_tunnel_net *itn; | |
96f5a846 | 144 | const struct iphdr *iph; |
88c7664f ACM |
145 | const int type = icmp_hdr(skb)->type; |
146 | const int code = icmp_hdr(skb)->code; | |
1da177e4 | 147 | struct ip_tunnel *t; |
1da177e4 | 148 | |
1da177e4 LT |
149 | switch (type) { |
150 | default: | |
151 | case ICMP_PARAMETERPROB: | |
9f57c67c | 152 | return; |
1da177e4 LT |
153 | |
154 | case ICMP_DEST_UNREACH: | |
155 | switch (code) { | |
156 | case ICMP_SR_FAILED: | |
157 | case ICMP_PORT_UNREACH: | |
158 | /* Impossible event. */ | |
9f57c67c | 159 | return; |
1da177e4 LT |
160 | default: |
161 | /* All others are translated to HOST_UNREACH. | |
162 | rfc2003 contains "deep thoughts" about NET_UNREACH, | |
163 | I believe they are just ether pollution. --ANK | |
164 | */ | |
165 | break; | |
166 | } | |
167 | break; | |
9f57c67c | 168 | |
1da177e4 LT |
169 | case ICMP_TIME_EXCEEDED: |
170 | if (code != ICMP_EXC_TTL) | |
9f57c67c | 171 | return; |
1da177e4 | 172 | break; |
55be7a9c DM |
173 | |
174 | case ICMP_REDIRECT: | |
175 | break; | |
1da177e4 LT |
176 | } |
177 | ||
bda7bb46 | 178 | if (tpi->proto == htons(ETH_P_TEB)) |
c5441932 PS |
179 | itn = net_generic(net, gre_tap_net_id); |
180 | else | |
181 | itn = net_generic(net, ipgre_net_id); | |
182 | ||
c0c0c50f | 183 | iph = (const struct iphdr *)(icmp_hdr(skb) + 1); |
bda7bb46 PS |
184 | t = ip_tunnel_lookup(itn, skb->dev->ifindex, tpi->flags, |
185 | iph->daddr, iph->saddr, tpi->key); | |
d2083287 | 186 | |
51456b29 | 187 | if (!t) |
9f57c67c | 188 | return; |
36393395 | 189 | |
36393395 | 190 | if (t->parms.iph.daddr == 0 || |
f97c1e0c | 191 | ipv4_is_multicast(t->parms.iph.daddr)) |
9f57c67c | 192 | return; |
1da177e4 LT |
193 | |
194 | if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED) | |
9f57c67c | 195 | return; |
1da177e4 | 196 | |
da6185d8 | 197 | if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO)) |
1da177e4 LT |
198 | t->err_count++; |
199 | else | |
200 | t->err_count = 1; | |
201 | t->err_time = jiffies; | |
9f57c67c PS |
202 | } |
203 | ||
204 | static void gre_err(struct sk_buff *skb, u32 info) | |
205 | { | |
206 | /* All the routers (except for Linux) return only | |
207 | * 8 bytes of packet payload. It means, that precise relaying of | |
208 | * ICMP in the real Internet is absolutely infeasible. | |
209 | * | |
210 | * Moreover, Cisco "wise men" put GRE key to the third word | |
211 | * in GRE header. It makes impossible maintaining even soft | |
212 | * state for keyed | |
213 | * GRE tunnels with enabled checksum. Tell them "thank you". | |
214 | * | |
215 | * Well, I wonder, rfc1812 was written by Cisco employee, | |
216 | * what the hell these idiots break standards established | |
217 | * by themselves??? | |
218 | */ | |
219 | ||
220 | const int type = icmp_hdr(skb)->type; | |
221 | const int code = icmp_hdr(skb)->code; | |
222 | struct tnl_ptk_info tpi; | |
223 | bool csum_err = false; | |
224 | ||
da73b4e9 | 225 | if (gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IP)) < 0) { |
9f57c67c PS |
226 | if (!csum_err) /* ignore csum errors. */ |
227 | return; | |
228 | } | |
229 | ||
230 | if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) { | |
231 | ipv4_update_pmtu(skb, dev_net(skb->dev), info, | |
232 | skb->dev->ifindex, 0, IPPROTO_GRE, 0); | |
233 | return; | |
234 | } | |
235 | if (type == ICMP_REDIRECT) { | |
236 | ipv4_redirect(skb, dev_net(skb->dev), skb->dev->ifindex, 0, | |
237 | IPPROTO_GRE, 0); | |
238 | return; | |
239 | } | |
240 | ||
241 | ipgre_err(skb, info, &tpi); | |
1da177e4 LT |
242 | } |
243 | ||
2e15ea39 PS |
244 | static __be64 key_to_tunnel_id(__be32 key) |
245 | { | |
246 | #ifdef __BIG_ENDIAN | |
247 | return (__force __be64)((__force u32)key); | |
248 | #else | |
249 | return (__force __be64)((__force u64)key << 32); | |
250 | #endif | |
251 | } | |
252 | ||
253 | /* Returns the least-significant 32 bits of a __be64. */ | |
254 | static __be32 tunnel_id_to_key(__be64 x) | |
255 | { | |
256 | #ifdef __BIG_ENDIAN | |
257 | return (__force __be32)x; | |
258 | #else | |
259 | return (__force __be32)((__force u64)x >> 32); | |
260 | #endif | |
261 | } | |
262 | ||
125372fa JB |
263 | static int __ipgre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi, |
264 | struct ip_tunnel_net *itn, int hdr_len, bool raw_proto) | |
1da177e4 | 265 | { |
2e15ea39 | 266 | struct metadata_dst *tun_dst = NULL; |
b71d1d42 | 267 | const struct iphdr *iph; |
1da177e4 | 268 | struct ip_tunnel *tunnel; |
1da177e4 | 269 | |
c5441932 | 270 | iph = ip_hdr(skb); |
bda7bb46 PS |
271 | tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, tpi->flags, |
272 | iph->saddr, iph->daddr, tpi->key); | |
e1a80002 | 273 | |
d2083287 | 274 | if (tunnel) { |
125372fa JB |
275 | if (__iptunnel_pull_header(skb, hdr_len, tpi->proto, |
276 | raw_proto, false) < 0) | |
244a797b JB |
277 | goto drop; |
278 | ||
e271c7b4 JB |
279 | if (tunnel->dev->type != ARPHRD_NONE) |
280 | skb_pop_mac_header(skb); | |
281 | else | |
282 | skb_reset_mac_header(skb); | |
2e15ea39 | 283 | if (tunnel->collect_md) { |
c29a70d2 PS |
284 | __be16 flags; |
285 | __be64 tun_id; | |
2e15ea39 | 286 | |
c29a70d2 PS |
287 | flags = tpi->flags & (TUNNEL_CSUM | TUNNEL_KEY); |
288 | tun_id = key_to_tunnel_id(tpi->key); | |
289 | tun_dst = ip_tun_rx_dst(skb, flags, tun_id, 0); | |
2e15ea39 PS |
290 | if (!tun_dst) |
291 | return PACKET_REJECT; | |
2e15ea39 PS |
292 | } |
293 | ||
294 | ip_tunnel_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error); | |
bda7bb46 | 295 | return PACKET_RCVD; |
1da177e4 | 296 | } |
125372fa | 297 | return PACKET_NEXT; |
244a797b JB |
298 | |
299 | drop: | |
300 | kfree_skb(skb); | |
301 | return PACKET_RCVD; | |
1da177e4 LT |
302 | } |
303 | ||
125372fa JB |
304 | static int ipgre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi, |
305 | int hdr_len) | |
306 | { | |
307 | struct net *net = dev_net(skb->dev); | |
308 | struct ip_tunnel_net *itn; | |
309 | int res; | |
310 | ||
311 | if (tpi->proto == htons(ETH_P_TEB)) | |
312 | itn = net_generic(net, gre_tap_net_id); | |
313 | else | |
314 | itn = net_generic(net, ipgre_net_id); | |
315 | ||
316 | res = __ipgre_rcv(skb, tpi, itn, hdr_len, false); | |
317 | if (res == PACKET_NEXT && tpi->proto == htons(ETH_P_TEB)) { | |
318 | /* ipgre tunnels in collect metadata mode should receive | |
319 | * also ETH_P_TEB traffic. | |
320 | */ | |
321 | itn = net_generic(net, ipgre_net_id); | |
322 | res = __ipgre_rcv(skb, tpi, itn, hdr_len, true); | |
323 | } | |
324 | return res; | |
325 | } | |
326 | ||
9f57c67c PS |
327 | static int gre_rcv(struct sk_buff *skb) |
328 | { | |
329 | struct tnl_ptk_info tpi; | |
330 | bool csum_err = false; | |
95f5c64c | 331 | int hdr_len; |
9f57c67c PS |
332 | |
333 | #ifdef CONFIG_NET_IPGRE_BROADCAST | |
334 | if (ipv4_is_multicast(ip_hdr(skb)->daddr)) { | |
335 | /* Looped back packet, drop it! */ | |
336 | if (rt_is_output_route(skb_rtable(skb))) | |
337 | goto drop; | |
338 | } | |
339 | #endif | |
340 | ||
da73b4e9 | 341 | hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IP)); |
f132ae7c | 342 | if (hdr_len < 0) |
95f5c64c TH |
343 | goto drop; |
344 | ||
244a797b | 345 | if (ipgre_rcv(skb, &tpi, hdr_len) == PACKET_RCVD) |
9f57c67c PS |
346 | return 0; |
347 | ||
348 | icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0); | |
349 | drop: | |
350 | kfree_skb(skb); | |
351 | return 0; | |
352 | } | |
353 | ||
c5441932 PS |
354 | static void __gre_xmit(struct sk_buff *skb, struct net_device *dev, |
355 | const struct iphdr *tnl_params, | |
356 | __be16 proto) | |
357 | { | |
358 | struct ip_tunnel *tunnel = netdev_priv(dev); | |
1da177e4 | 359 | |
c5441932 PS |
360 | if (tunnel->parms.o_flags & TUNNEL_SEQ) |
361 | tunnel->o_seqno++; | |
1da177e4 | 362 | |
c5441932 | 363 | /* Push GRE header. */ |
182a352d TH |
364 | gre_build_header(skb, tunnel->tun_hlen, |
365 | tunnel->parms.o_flags, proto, tunnel->parms.o_key, | |
366 | htonl(tunnel->o_seqno)); | |
54bc9bac | 367 | |
2e15ea39 | 368 | skb_set_inner_protocol(skb, proto); |
bf3d6a8f | 369 | ip_tunnel_xmit(skb, dev, tnl_params, tnl_params->protocol); |
c5441932 | 370 | } |
1da177e4 | 371 | |
aed069df | 372 | static int gre_handle_offloads(struct sk_buff *skb, bool csum) |
b2acd1dc | 373 | { |
6fa79666 | 374 | return iptunnel_handle_offloads(skb, csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE); |
b2acd1dc PS |
375 | } |
376 | ||
fc4099f1 PS |
377 | static struct rtable *gre_get_rt(struct sk_buff *skb, |
378 | struct net_device *dev, | |
379 | struct flowi4 *fl, | |
380 | const struct ip_tunnel_key *key) | |
381 | { | |
382 | struct net *net = dev_net(dev); | |
383 | ||
384 | memset(fl, 0, sizeof(*fl)); | |
385 | fl->daddr = key->u.ipv4.dst; | |
386 | fl->saddr = key->u.ipv4.src; | |
387 | fl->flowi4_tos = RT_TOS(key->tos); | |
388 | fl->flowi4_mark = skb->mark; | |
389 | fl->flowi4_proto = IPPROTO_GRE; | |
390 | ||
391 | return ip_route_output_key(net, fl); | |
392 | } | |
393 | ||
2090714e JB |
394 | static void gre_fb_xmit(struct sk_buff *skb, struct net_device *dev, |
395 | __be16 proto) | |
2e15ea39 PS |
396 | { |
397 | struct ip_tunnel_info *tun_info; | |
2e15ea39 | 398 | const struct ip_tunnel_key *key; |
db3c6139 | 399 | struct rtable *rt = NULL; |
2e15ea39 | 400 | struct flowi4 fl; |
2e15ea39 PS |
401 | int min_headroom; |
402 | int tunnel_hlen; | |
403 | __be16 df, flags; | |
db3c6139 | 404 | bool use_cache; |
2e15ea39 PS |
405 | int err; |
406 | ||
61adedf3 | 407 | tun_info = skb_tunnel_info(skb); |
7f9562a1 JB |
408 | if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) || |
409 | ip_tunnel_info_af(tun_info) != AF_INET)) | |
2e15ea39 PS |
410 | goto err_free_skb; |
411 | ||
412 | key = &tun_info->key; | |
db3c6139 DB |
413 | use_cache = ip_tunnel_dst_cache_usable(skb, tun_info); |
414 | if (use_cache) | |
415 | rt = dst_cache_get_ip4(&tun_info->dst_cache, &fl.saddr); | |
3c1cb4d2 PA |
416 | if (!rt) { |
417 | rt = gre_get_rt(skb, dev, &fl, key); | |
418 | if (IS_ERR(rt)) | |
419 | goto err_free_skb; | |
db3c6139 | 420 | if (use_cache) |
3c1cb4d2 PA |
421 | dst_cache_set_ip4(&tun_info->dst_cache, &rt->dst, |
422 | fl.saddr); | |
423 | } | |
2e15ea39 | 424 | |
95f5c64c | 425 | tunnel_hlen = gre_calc_hlen(key->tun_flags); |
2e15ea39 PS |
426 | |
427 | min_headroom = LL_RESERVED_SPACE(rt->dst.dev) + rt->dst.header_len | |
428 | + tunnel_hlen + sizeof(struct iphdr); | |
429 | if (skb_headroom(skb) < min_headroom || skb_header_cloned(skb)) { | |
430 | int head_delta = SKB_DATA_ALIGN(min_headroom - | |
431 | skb_headroom(skb) + | |
432 | 16); | |
433 | err = pskb_expand_head(skb, max_t(int, head_delta, 0), | |
434 | 0, GFP_ATOMIC); | |
435 | if (unlikely(err)) | |
436 | goto err_free_rt; | |
437 | } | |
438 | ||
439 | /* Push Tunnel header. */ | |
aed069df | 440 | if (gre_handle_offloads(skb, !!(tun_info->key.tun_flags & TUNNEL_CSUM))) |
2e15ea39 | 441 | goto err_free_rt; |
2e15ea39 PS |
442 | |
443 | flags = tun_info->key.tun_flags & (TUNNEL_CSUM | TUNNEL_KEY); | |
cba65321 | 444 | gre_build_header(skb, tunnel_hlen, flags, proto, |
182a352d | 445 | tunnel_id_to_key(tun_info->key.tun_id), 0); |
2e15ea39 PS |
446 | |
447 | df = key->tun_flags & TUNNEL_DONT_FRAGMENT ? htons(IP_DF) : 0; | |
039f5062 PS |
448 | |
449 | iptunnel_xmit(skb->sk, rt, skb, fl.saddr, key->u.ipv4.dst, IPPROTO_GRE, | |
450 | key->tos, key->ttl, df, false); | |
2e15ea39 PS |
451 | return; |
452 | ||
453 | err_free_rt: | |
454 | ip_rt_put(rt); | |
455 | err_free_skb: | |
456 | kfree_skb(skb); | |
457 | dev->stats.tx_dropped++; | |
458 | } | |
459 | ||
fc4099f1 PS |
460 | static int gre_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb) |
461 | { | |
462 | struct ip_tunnel_info *info = skb_tunnel_info(skb); | |
463 | struct rtable *rt; | |
464 | struct flowi4 fl4; | |
465 | ||
466 | if (ip_tunnel_info_af(info) != AF_INET) | |
467 | return -EINVAL; | |
468 | ||
469 | rt = gre_get_rt(skb, dev, &fl4, &info->key); | |
470 | if (IS_ERR(rt)) | |
471 | return PTR_ERR(rt); | |
472 | ||
473 | ip_rt_put(rt); | |
474 | info->key.u.ipv4.src = fl4.saddr; | |
475 | return 0; | |
476 | } | |
477 | ||
c5441932 PS |
478 | static netdev_tx_t ipgre_xmit(struct sk_buff *skb, |
479 | struct net_device *dev) | |
480 | { | |
481 | struct ip_tunnel *tunnel = netdev_priv(dev); | |
482 | const struct iphdr *tnl_params; | |
1da177e4 | 483 | |
2e15ea39 | 484 | if (tunnel->collect_md) { |
2090714e | 485 | gre_fb_xmit(skb, dev, skb->protocol); |
2e15ea39 PS |
486 | return NETDEV_TX_OK; |
487 | } | |
488 | ||
c5441932 PS |
489 | if (dev->header_ops) { |
490 | /* Need space for new headers */ | |
491 | if (skb_cow_head(skb, dev->needed_headroom - | |
2bac7cb3 | 492 | (tunnel->hlen + sizeof(struct iphdr)))) |
c5441932 | 493 | goto free_skb; |
1da177e4 | 494 | |
c5441932 | 495 | tnl_params = (const struct iphdr *)skb->data; |
1da177e4 | 496 | |
c5441932 PS |
497 | /* Pull skb since ip_tunnel_xmit() needs skb->data pointing |
498 | * to gre header. | |
499 | */ | |
500 | skb_pull(skb, tunnel->hlen + sizeof(struct iphdr)); | |
8a0033a9 | 501 | skb_reset_mac_header(skb); |
c5441932 PS |
502 | } else { |
503 | if (skb_cow_head(skb, dev->needed_headroom)) | |
504 | goto free_skb; | |
1da177e4 | 505 | |
c5441932 | 506 | tnl_params = &tunnel->parms.iph; |
1da177e4 LT |
507 | } |
508 | ||
aed069df AD |
509 | if (gre_handle_offloads(skb, !!(tunnel->parms.o_flags & TUNNEL_CSUM))) |
510 | goto free_skb; | |
8a0033a9 | 511 | |
c5441932 | 512 | __gre_xmit(skb, dev, tnl_params, skb->protocol); |
6ed10654 | 513 | return NETDEV_TX_OK; |
1da177e4 | 514 | |
c5441932 | 515 | free_skb: |
3acfa1e7 | 516 | kfree_skb(skb); |
c5441932 | 517 | dev->stats.tx_dropped++; |
6ed10654 | 518 | return NETDEV_TX_OK; |
1da177e4 LT |
519 | } |
520 | ||
c5441932 PS |
521 | static netdev_tx_t gre_tap_xmit(struct sk_buff *skb, |
522 | struct net_device *dev) | |
ee34c1eb | 523 | { |
c5441932 | 524 | struct ip_tunnel *tunnel = netdev_priv(dev); |
ee34c1eb | 525 | |
2e15ea39 | 526 | if (tunnel->collect_md) { |
2090714e | 527 | gre_fb_xmit(skb, dev, htons(ETH_P_TEB)); |
2e15ea39 PS |
528 | return NETDEV_TX_OK; |
529 | } | |
530 | ||
aed069df AD |
531 | if (gre_handle_offloads(skb, !!(tunnel->parms.o_flags & TUNNEL_CSUM))) |
532 | goto free_skb; | |
ee34c1eb | 533 | |
c5441932 PS |
534 | if (skb_cow_head(skb, dev->needed_headroom)) |
535 | goto free_skb; | |
42aa9162 | 536 | |
c5441932 | 537 | __gre_xmit(skb, dev, &tunnel->parms.iph, htons(ETH_P_TEB)); |
c5441932 | 538 | return NETDEV_TX_OK; |
ee34c1eb | 539 | |
c5441932 | 540 | free_skb: |
3acfa1e7 | 541 | kfree_skb(skb); |
c5441932 PS |
542 | dev->stats.tx_dropped++; |
543 | return NETDEV_TX_OK; | |
ee34c1eb MS |
544 | } |
545 | ||
c5441932 PS |
546 | static int ipgre_tunnel_ioctl(struct net_device *dev, |
547 | struct ifreq *ifr, int cmd) | |
1da177e4 | 548 | { |
4565e991 | 549 | int err; |
1da177e4 | 550 | struct ip_tunnel_parm p; |
1da177e4 | 551 | |
c5441932 PS |
552 | if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) |
553 | return -EFAULT; | |
6c734fb8 CW |
554 | if (cmd == SIOCADDTUNNEL || cmd == SIOCCHGTUNNEL) { |
555 | if (p.iph.version != 4 || p.iph.protocol != IPPROTO_GRE || | |
556 | p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)) || | |
557 | ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))) | |
558 | return -EINVAL; | |
1da177e4 | 559 | } |
c5441932 PS |
560 | p.i_flags = gre_flags_to_tnl_flags(p.i_flags); |
561 | p.o_flags = gre_flags_to_tnl_flags(p.o_flags); | |
1da177e4 | 562 | |
c5441932 PS |
563 | err = ip_tunnel_ioctl(dev, &p, cmd); |
564 | if (err) | |
565 | return err; | |
1da177e4 | 566 | |
95f5c64c TH |
567 | p.i_flags = gre_tnl_flags_to_gre_flags(p.i_flags); |
568 | p.o_flags = gre_tnl_flags_to_gre_flags(p.o_flags); | |
c5441932 PS |
569 | |
570 | if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) | |
571 | return -EFAULT; | |
1da177e4 LT |
572 | return 0; |
573 | } | |
574 | ||
1da177e4 LT |
575 | /* Nice toy. Unfortunately, useless in real life :-) |
576 | It allows to construct virtual multiprotocol broadcast "LAN" | |
577 | over the Internet, provided multicast routing is tuned. | |
578 | ||
579 | ||
580 | I have no idea was this bicycle invented before me, | |
581 | so that I had to set ARPHRD_IPGRE to a random value. | |
582 | I have an impression, that Cisco could make something similar, | |
583 | but this feature is apparently missing in IOS<=11.2(8). | |
e905a9ed | 584 | |
1da177e4 LT |
585 | I set up 10.66.66/24 and fec0:6666:6666::0/96 as virtual networks |
586 | with broadcast 224.66.66.66. If you have access to mbone, play with me :-) | |
587 | ||
588 | ping -t 255 224.66.66.66 | |
589 | ||
590 | If nobody answers, mbone does not work. | |
591 | ||
592 | ip tunnel add Universe mode gre remote 224.66.66.66 local <Your_real_addr> ttl 255 | |
593 | ip addr add 10.66.66.<somewhat>/24 dev Universe | |
594 | ifconfig Universe up | |
595 | ifconfig Universe add fe80::<Your_real_addr>/10 | |
596 | ifconfig Universe add fec0:6666:6666::<Your_real_addr>/96 | |
597 | ftp 10.66.66.66 | |
598 | ... | |
599 | ftp fec0:6666:6666::193.233.7.65 | |
600 | ... | |
1da177e4 | 601 | */ |
3b04ddde SH |
602 | static int ipgre_header(struct sk_buff *skb, struct net_device *dev, |
603 | unsigned short type, | |
1507850b | 604 | const void *daddr, const void *saddr, unsigned int len) |
1da177e4 | 605 | { |
2941a486 | 606 | struct ip_tunnel *t = netdev_priv(dev); |
c5441932 PS |
607 | struct iphdr *iph; |
608 | struct gre_base_hdr *greh; | |
1da177e4 | 609 | |
c5441932 PS |
610 | iph = (struct iphdr *)skb_push(skb, t->hlen + sizeof(*iph)); |
611 | greh = (struct gre_base_hdr *)(iph+1); | |
95f5c64c | 612 | greh->flags = gre_tnl_flags_to_gre_flags(t->parms.o_flags); |
c5441932 | 613 | greh->protocol = htons(type); |
1da177e4 | 614 | |
c5441932 | 615 | memcpy(iph, &t->parms.iph, sizeof(struct iphdr)); |
e905a9ed | 616 | |
c5441932 | 617 | /* Set the source hardware address. */ |
1da177e4 LT |
618 | if (saddr) |
619 | memcpy(&iph->saddr, saddr, 4); | |
6d55cb91 | 620 | if (daddr) |
1da177e4 | 621 | memcpy(&iph->daddr, daddr, 4); |
6d55cb91 | 622 | if (iph->daddr) |
77a482bd | 623 | return t->hlen + sizeof(*iph); |
e905a9ed | 624 | |
c5441932 | 625 | return -(t->hlen + sizeof(*iph)); |
1da177e4 LT |
626 | } |
627 | ||
6a5f44d7 TT |
628 | static int ipgre_header_parse(const struct sk_buff *skb, unsigned char *haddr) |
629 | { | |
b71d1d42 | 630 | const struct iphdr *iph = (const struct iphdr *) skb_mac_header(skb); |
6a5f44d7 TT |
631 | memcpy(haddr, &iph->saddr, 4); |
632 | return 4; | |
633 | } | |
634 | ||
3b04ddde SH |
635 | static const struct header_ops ipgre_header_ops = { |
636 | .create = ipgre_header, | |
6a5f44d7 | 637 | .parse = ipgre_header_parse, |
3b04ddde SH |
638 | }; |
639 | ||
6a5f44d7 | 640 | #ifdef CONFIG_NET_IPGRE_BROADCAST |
1da177e4 LT |
641 | static int ipgre_open(struct net_device *dev) |
642 | { | |
2941a486 | 643 | struct ip_tunnel *t = netdev_priv(dev); |
1da177e4 | 644 | |
f97c1e0c | 645 | if (ipv4_is_multicast(t->parms.iph.daddr)) { |
cbb1e85f DM |
646 | struct flowi4 fl4; |
647 | struct rtable *rt; | |
648 | ||
b57708ad | 649 | rt = ip_route_output_gre(t->net, &fl4, |
cbb1e85f DM |
650 | t->parms.iph.daddr, |
651 | t->parms.iph.saddr, | |
652 | t->parms.o_key, | |
653 | RT_TOS(t->parms.iph.tos), | |
654 | t->parms.link); | |
b23dd4fe | 655 | if (IS_ERR(rt)) |
1da177e4 | 656 | return -EADDRNOTAVAIL; |
d8d1f30b | 657 | dev = rt->dst.dev; |
1da177e4 | 658 | ip_rt_put(rt); |
51456b29 | 659 | if (!__in_dev_get_rtnl(dev)) |
1da177e4 LT |
660 | return -EADDRNOTAVAIL; |
661 | t->mlink = dev->ifindex; | |
e5ed6399 | 662 | ip_mc_inc_group(__in_dev_get_rtnl(dev), t->parms.iph.daddr); |
1da177e4 LT |
663 | } |
664 | return 0; | |
665 | } | |
666 | ||
667 | static int ipgre_close(struct net_device *dev) | |
668 | { | |
2941a486 | 669 | struct ip_tunnel *t = netdev_priv(dev); |
b8c26a33 | 670 | |
f97c1e0c | 671 | if (ipv4_is_multicast(t->parms.iph.daddr) && t->mlink) { |
7fee0ca2 | 672 | struct in_device *in_dev; |
b57708ad | 673 | in_dev = inetdev_by_index(t->net, t->mlink); |
8723e1b4 | 674 | if (in_dev) |
1da177e4 | 675 | ip_mc_dec_group(in_dev, t->parms.iph.daddr); |
1da177e4 LT |
676 | } |
677 | return 0; | |
678 | } | |
1da177e4 LT |
679 | #endif |
680 | ||
b8c26a33 SH |
681 | static const struct net_device_ops ipgre_netdev_ops = { |
682 | .ndo_init = ipgre_tunnel_init, | |
c5441932 | 683 | .ndo_uninit = ip_tunnel_uninit, |
b8c26a33 SH |
684 | #ifdef CONFIG_NET_IPGRE_BROADCAST |
685 | .ndo_open = ipgre_open, | |
686 | .ndo_stop = ipgre_close, | |
687 | #endif | |
c5441932 | 688 | .ndo_start_xmit = ipgre_xmit, |
b8c26a33 | 689 | .ndo_do_ioctl = ipgre_tunnel_ioctl, |
c5441932 PS |
690 | .ndo_change_mtu = ip_tunnel_change_mtu, |
691 | .ndo_get_stats64 = ip_tunnel_get_stats64, | |
1e99584b | 692 | .ndo_get_iflink = ip_tunnel_get_iflink, |
b8c26a33 SH |
693 | }; |
694 | ||
6b78f16e ED |
695 | #define GRE_FEATURES (NETIF_F_SG | \ |
696 | NETIF_F_FRAGLIST | \ | |
697 | NETIF_F_HIGHDMA | \ | |
698 | NETIF_F_HW_CSUM) | |
699 | ||
1da177e4 LT |
700 | static void ipgre_tunnel_setup(struct net_device *dev) |
701 | { | |
b8c26a33 | 702 | dev->netdev_ops = &ipgre_netdev_ops; |
5a455275 | 703 | dev->type = ARPHRD_IPGRE; |
c5441932 PS |
704 | ip_tunnel_setup(dev, ipgre_net_id); |
705 | } | |
1da177e4 | 706 | |
c5441932 PS |
707 | static void __gre_tunnel_init(struct net_device *dev) |
708 | { | |
709 | struct ip_tunnel *tunnel; | |
4565e991 | 710 | int t_hlen; |
c5441932 PS |
711 | |
712 | tunnel = netdev_priv(dev); | |
95f5c64c | 713 | tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags); |
c5441932 PS |
714 | tunnel->parms.iph.protocol = IPPROTO_GRE; |
715 | ||
4565e991 TH |
716 | tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen; |
717 | ||
718 | t_hlen = tunnel->hlen + sizeof(struct iphdr); | |
719 | ||
720 | dev->needed_headroom = LL_MAX_HEADER + t_hlen + 4; | |
721 | dev->mtu = ETH_DATA_LEN - t_hlen - 4; | |
6b78f16e | 722 | |
b57708ad | 723 | dev->features |= GRE_FEATURES; |
6b78f16e | 724 | dev->hw_features |= GRE_FEATURES; |
c5441932 PS |
725 | |
726 | if (!(tunnel->parms.o_flags & TUNNEL_SEQ)) { | |
a0ca153f AD |
727 | /* TCP offload with GRE SEQ is not supported, nor |
728 | * can we support 2 levels of outer headers requiring | |
729 | * an update. | |
730 | */ | |
731 | if (!(tunnel->parms.o_flags & TUNNEL_CSUM) || | |
732 | (tunnel->encap.type == TUNNEL_ENCAP_NONE)) { | |
733 | dev->features |= NETIF_F_GSO_SOFTWARE; | |
734 | dev->hw_features |= NETIF_F_GSO_SOFTWARE; | |
735 | } | |
736 | ||
c5441932 PS |
737 | /* Can use a lockless transmit, unless we generate |
738 | * output sequences | |
739 | */ | |
740 | dev->features |= NETIF_F_LLTX; | |
741 | } | |
1da177e4 LT |
742 | } |
743 | ||
744 | static int ipgre_tunnel_init(struct net_device *dev) | |
745 | { | |
c5441932 PS |
746 | struct ip_tunnel *tunnel = netdev_priv(dev); |
747 | struct iphdr *iph = &tunnel->parms.iph; | |
1da177e4 | 748 | |
c5441932 | 749 | __gre_tunnel_init(dev); |
1da177e4 | 750 | |
c5441932 PS |
751 | memcpy(dev->dev_addr, &iph->saddr, 4); |
752 | memcpy(dev->broadcast, &iph->daddr, 4); | |
1da177e4 | 753 | |
c5441932 | 754 | dev->flags = IFF_NOARP; |
02875878 | 755 | netif_keep_dst(dev); |
c5441932 | 756 | dev->addr_len = 4; |
1da177e4 | 757 | |
a64b04d8 | 758 | if (iph->daddr && !tunnel->collect_md) { |
1da177e4 | 759 | #ifdef CONFIG_NET_IPGRE_BROADCAST |
f97c1e0c | 760 | if (ipv4_is_multicast(iph->daddr)) { |
1da177e4 LT |
761 | if (!iph->saddr) |
762 | return -EINVAL; | |
763 | dev->flags = IFF_BROADCAST; | |
3b04ddde | 764 | dev->header_ops = &ipgre_header_ops; |
1da177e4 LT |
765 | } |
766 | #endif | |
a64b04d8 | 767 | } else if (!tunnel->collect_md) { |
6a5f44d7 | 768 | dev->header_ops = &ipgre_header_ops; |
a64b04d8 | 769 | } |
1da177e4 | 770 | |
c5441932 | 771 | return ip_tunnel_init(dev); |
1da177e4 LT |
772 | } |
773 | ||
9f57c67c PS |
774 | static const struct gre_protocol ipgre_protocol = { |
775 | .handler = gre_rcv, | |
776 | .err_handler = gre_err, | |
1da177e4 LT |
777 | }; |
778 | ||
2c8c1e72 | 779 | static int __net_init ipgre_init_net(struct net *net) |
59a4c759 | 780 | { |
c5441932 | 781 | return ip_tunnel_init_net(net, ipgre_net_id, &ipgre_link_ops, NULL); |
59a4c759 PE |
782 | } |
783 | ||
2c8c1e72 | 784 | static void __net_exit ipgre_exit_net(struct net *net) |
59a4c759 | 785 | { |
c5441932 | 786 | struct ip_tunnel_net *itn = net_generic(net, ipgre_net_id); |
6c742e71 | 787 | ip_tunnel_delete_net(itn, &ipgre_link_ops); |
59a4c759 PE |
788 | } |
789 | ||
790 | static struct pernet_operations ipgre_net_ops = { | |
791 | .init = ipgre_init_net, | |
792 | .exit = ipgre_exit_net, | |
cfb8fbf2 | 793 | .id = &ipgre_net_id, |
c5441932 | 794 | .size = sizeof(struct ip_tunnel_net), |
59a4c759 | 795 | }; |
1da177e4 | 796 | |
c19e654d HX |
797 | static int ipgre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[]) |
798 | { | |
799 | __be16 flags; | |
800 | ||
801 | if (!data) | |
802 | return 0; | |
803 | ||
804 | flags = 0; | |
805 | if (data[IFLA_GRE_IFLAGS]) | |
806 | flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]); | |
807 | if (data[IFLA_GRE_OFLAGS]) | |
808 | flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]); | |
809 | if (flags & (GRE_VERSION|GRE_ROUTING)) | |
810 | return -EINVAL; | |
811 | ||
946b636f JB |
812 | if (data[IFLA_GRE_COLLECT_METADATA] && |
813 | data[IFLA_GRE_ENCAP_TYPE] && | |
814 | nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]) != TUNNEL_ENCAP_NONE) | |
815 | return -EINVAL; | |
816 | ||
c19e654d HX |
817 | return 0; |
818 | } | |
819 | ||
e1a80002 HX |
820 | static int ipgre_tap_validate(struct nlattr *tb[], struct nlattr *data[]) |
821 | { | |
822 | __be32 daddr; | |
823 | ||
824 | if (tb[IFLA_ADDRESS]) { | |
825 | if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) | |
826 | return -EINVAL; | |
827 | if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) | |
828 | return -EADDRNOTAVAIL; | |
829 | } | |
830 | ||
831 | if (!data) | |
832 | goto out; | |
833 | ||
834 | if (data[IFLA_GRE_REMOTE]) { | |
835 | memcpy(&daddr, nla_data(data[IFLA_GRE_REMOTE]), 4); | |
836 | if (!daddr) | |
837 | return -EINVAL; | |
838 | } | |
839 | ||
840 | out: | |
841 | return ipgre_tunnel_validate(tb, data); | |
842 | } | |
843 | ||
2e15ea39 PS |
844 | static void ipgre_netlink_parms(struct net_device *dev, |
845 | struct nlattr *data[], | |
846 | struct nlattr *tb[], | |
847 | struct ip_tunnel_parm *parms) | |
c19e654d | 848 | { |
7bb82d92 | 849 | memset(parms, 0, sizeof(*parms)); |
c19e654d HX |
850 | |
851 | parms->iph.protocol = IPPROTO_GRE; | |
852 | ||
853 | if (!data) | |
854 | return; | |
855 | ||
856 | if (data[IFLA_GRE_LINK]) | |
857 | parms->link = nla_get_u32(data[IFLA_GRE_LINK]); | |
858 | ||
859 | if (data[IFLA_GRE_IFLAGS]) | |
c5441932 | 860 | parms->i_flags = gre_flags_to_tnl_flags(nla_get_be16(data[IFLA_GRE_IFLAGS])); |
c19e654d HX |
861 | |
862 | if (data[IFLA_GRE_OFLAGS]) | |
c5441932 | 863 | parms->o_flags = gre_flags_to_tnl_flags(nla_get_be16(data[IFLA_GRE_OFLAGS])); |
c19e654d HX |
864 | |
865 | if (data[IFLA_GRE_IKEY]) | |
866 | parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]); | |
867 | ||
868 | if (data[IFLA_GRE_OKEY]) | |
869 | parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]); | |
870 | ||
871 | if (data[IFLA_GRE_LOCAL]) | |
67b61f6c | 872 | parms->iph.saddr = nla_get_in_addr(data[IFLA_GRE_LOCAL]); |
c19e654d HX |
873 | |
874 | if (data[IFLA_GRE_REMOTE]) | |
67b61f6c | 875 | parms->iph.daddr = nla_get_in_addr(data[IFLA_GRE_REMOTE]); |
c19e654d HX |
876 | |
877 | if (data[IFLA_GRE_TTL]) | |
878 | parms->iph.ttl = nla_get_u8(data[IFLA_GRE_TTL]); | |
879 | ||
880 | if (data[IFLA_GRE_TOS]) | |
881 | parms->iph.tos = nla_get_u8(data[IFLA_GRE_TOS]); | |
882 | ||
883 | if (!data[IFLA_GRE_PMTUDISC] || nla_get_u8(data[IFLA_GRE_PMTUDISC])) | |
884 | parms->iph.frag_off = htons(IP_DF); | |
2e15ea39 PS |
885 | |
886 | if (data[IFLA_GRE_COLLECT_METADATA]) { | |
887 | struct ip_tunnel *t = netdev_priv(dev); | |
888 | ||
889 | t->collect_md = true; | |
e271c7b4 JB |
890 | if (dev->type == ARPHRD_IPGRE) |
891 | dev->type = ARPHRD_NONE; | |
2e15ea39 | 892 | } |
c19e654d HX |
893 | } |
894 | ||
4565e991 TH |
895 | /* This function returns true when ENCAP attributes are present in the nl msg */ |
896 | static bool ipgre_netlink_encap_parms(struct nlattr *data[], | |
897 | struct ip_tunnel_encap *ipencap) | |
898 | { | |
899 | bool ret = false; | |
900 | ||
901 | memset(ipencap, 0, sizeof(*ipencap)); | |
902 | ||
903 | if (!data) | |
904 | return ret; | |
905 | ||
906 | if (data[IFLA_GRE_ENCAP_TYPE]) { | |
907 | ret = true; | |
908 | ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]); | |
909 | } | |
910 | ||
911 | if (data[IFLA_GRE_ENCAP_FLAGS]) { | |
912 | ret = true; | |
913 | ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]); | |
914 | } | |
915 | ||
916 | if (data[IFLA_GRE_ENCAP_SPORT]) { | |
917 | ret = true; | |
3e97fa70 | 918 | ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]); |
4565e991 TH |
919 | } |
920 | ||
921 | if (data[IFLA_GRE_ENCAP_DPORT]) { | |
922 | ret = true; | |
3e97fa70 | 923 | ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]); |
4565e991 TH |
924 | } |
925 | ||
926 | return ret; | |
927 | } | |
928 | ||
c5441932 | 929 | static int gre_tap_init(struct net_device *dev) |
e1a80002 | 930 | { |
c5441932 | 931 | __gre_tunnel_init(dev); |
bec94d43 | 932 | dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; |
e1a80002 | 933 | |
c5441932 | 934 | return ip_tunnel_init(dev); |
e1a80002 HX |
935 | } |
936 | ||
c5441932 PS |
937 | static const struct net_device_ops gre_tap_netdev_ops = { |
938 | .ndo_init = gre_tap_init, | |
939 | .ndo_uninit = ip_tunnel_uninit, | |
940 | .ndo_start_xmit = gre_tap_xmit, | |
b8c26a33 SH |
941 | .ndo_set_mac_address = eth_mac_addr, |
942 | .ndo_validate_addr = eth_validate_addr, | |
c5441932 PS |
943 | .ndo_change_mtu = ip_tunnel_change_mtu, |
944 | .ndo_get_stats64 = ip_tunnel_get_stats64, | |
1e99584b | 945 | .ndo_get_iflink = ip_tunnel_get_iflink, |
fc4099f1 | 946 | .ndo_fill_metadata_dst = gre_fill_metadata_dst, |
b8c26a33 SH |
947 | }; |
948 | ||
e1a80002 HX |
949 | static void ipgre_tap_setup(struct net_device *dev) |
950 | { | |
e1a80002 | 951 | ether_setup(dev); |
d13b161c JB |
952 | dev->netdev_ops = &gre_tap_netdev_ops; |
953 | dev->priv_flags &= ~IFF_TX_SKB_SHARING; | |
954 | dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; | |
c5441932 | 955 | ip_tunnel_setup(dev, gre_tap_net_id); |
e1a80002 HX |
956 | } |
957 | ||
c5441932 PS |
958 | static int ipgre_newlink(struct net *src_net, struct net_device *dev, |
959 | struct nlattr *tb[], struct nlattr *data[]) | |
c19e654d | 960 | { |
c5441932 | 961 | struct ip_tunnel_parm p; |
4565e991 TH |
962 | struct ip_tunnel_encap ipencap; |
963 | ||
964 | if (ipgre_netlink_encap_parms(data, &ipencap)) { | |
965 | struct ip_tunnel *t = netdev_priv(dev); | |
966 | int err = ip_tunnel_encap_setup(t, &ipencap); | |
967 | ||
968 | if (err < 0) | |
969 | return err; | |
970 | } | |
c19e654d | 971 | |
2e15ea39 | 972 | ipgre_netlink_parms(dev, data, tb, &p); |
c5441932 | 973 | return ip_tunnel_newlink(dev, tb, &p); |
c19e654d HX |
974 | } |
975 | ||
976 | static int ipgre_changelink(struct net_device *dev, struct nlattr *tb[], | |
977 | struct nlattr *data[]) | |
978 | { | |
c19e654d | 979 | struct ip_tunnel_parm p; |
4565e991 TH |
980 | struct ip_tunnel_encap ipencap; |
981 | ||
982 | if (ipgre_netlink_encap_parms(data, &ipencap)) { | |
983 | struct ip_tunnel *t = netdev_priv(dev); | |
984 | int err = ip_tunnel_encap_setup(t, &ipencap); | |
985 | ||
986 | if (err < 0) | |
987 | return err; | |
988 | } | |
c19e654d | 989 | |
2e15ea39 | 990 | ipgre_netlink_parms(dev, data, tb, &p); |
c5441932 | 991 | return ip_tunnel_changelink(dev, tb, &p); |
c19e654d HX |
992 | } |
993 | ||
994 | static size_t ipgre_get_size(const struct net_device *dev) | |
995 | { | |
996 | return | |
997 | /* IFLA_GRE_LINK */ | |
998 | nla_total_size(4) + | |
999 | /* IFLA_GRE_IFLAGS */ | |
1000 | nla_total_size(2) + | |
1001 | /* IFLA_GRE_OFLAGS */ | |
1002 | nla_total_size(2) + | |
1003 | /* IFLA_GRE_IKEY */ | |
1004 | nla_total_size(4) + | |
1005 | /* IFLA_GRE_OKEY */ | |
1006 | nla_total_size(4) + | |
1007 | /* IFLA_GRE_LOCAL */ | |
1008 | nla_total_size(4) + | |
1009 | /* IFLA_GRE_REMOTE */ | |
1010 | nla_total_size(4) + | |
1011 | /* IFLA_GRE_TTL */ | |
1012 | nla_total_size(1) + | |
1013 | /* IFLA_GRE_TOS */ | |
1014 | nla_total_size(1) + | |
1015 | /* IFLA_GRE_PMTUDISC */ | |
1016 | nla_total_size(1) + | |
4565e991 TH |
1017 | /* IFLA_GRE_ENCAP_TYPE */ |
1018 | nla_total_size(2) + | |
1019 | /* IFLA_GRE_ENCAP_FLAGS */ | |
1020 | nla_total_size(2) + | |
1021 | /* IFLA_GRE_ENCAP_SPORT */ | |
1022 | nla_total_size(2) + | |
1023 | /* IFLA_GRE_ENCAP_DPORT */ | |
1024 | nla_total_size(2) + | |
2e15ea39 PS |
1025 | /* IFLA_GRE_COLLECT_METADATA */ |
1026 | nla_total_size(0) + | |
c19e654d HX |
1027 | 0; |
1028 | } | |
1029 | ||
1030 | static int ipgre_fill_info(struct sk_buff *skb, const struct net_device *dev) | |
1031 | { | |
1032 | struct ip_tunnel *t = netdev_priv(dev); | |
1033 | struct ip_tunnel_parm *p = &t->parms; | |
1034 | ||
f3756b79 | 1035 | if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) || |
95f5c64c TH |
1036 | nla_put_be16(skb, IFLA_GRE_IFLAGS, |
1037 | gre_tnl_flags_to_gre_flags(p->i_flags)) || | |
1038 | nla_put_be16(skb, IFLA_GRE_OFLAGS, | |
1039 | gre_tnl_flags_to_gre_flags(p->o_flags)) || | |
f3756b79 DM |
1040 | nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) || |
1041 | nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) || | |
930345ea JB |
1042 | nla_put_in_addr(skb, IFLA_GRE_LOCAL, p->iph.saddr) || |
1043 | nla_put_in_addr(skb, IFLA_GRE_REMOTE, p->iph.daddr) || | |
f3756b79 DM |
1044 | nla_put_u8(skb, IFLA_GRE_TTL, p->iph.ttl) || |
1045 | nla_put_u8(skb, IFLA_GRE_TOS, p->iph.tos) || | |
1046 | nla_put_u8(skb, IFLA_GRE_PMTUDISC, | |
1047 | !!(p->iph.frag_off & htons(IP_DF)))) | |
1048 | goto nla_put_failure; | |
4565e991 TH |
1049 | |
1050 | if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE, | |
1051 | t->encap.type) || | |
3e97fa70 SD |
1052 | nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT, |
1053 | t->encap.sport) || | |
1054 | nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT, | |
1055 | t->encap.dport) || | |
4565e991 | 1056 | nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS, |
e1b2cb65 | 1057 | t->encap.flags)) |
4565e991 TH |
1058 | goto nla_put_failure; |
1059 | ||
2e15ea39 PS |
1060 | if (t->collect_md) { |
1061 | if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA)) | |
1062 | goto nla_put_failure; | |
1063 | } | |
1064 | ||
c19e654d HX |
1065 | return 0; |
1066 | ||
1067 | nla_put_failure: | |
1068 | return -EMSGSIZE; | |
1069 | } | |
1070 | ||
1071 | static const struct nla_policy ipgre_policy[IFLA_GRE_MAX + 1] = { | |
1072 | [IFLA_GRE_LINK] = { .type = NLA_U32 }, | |
1073 | [IFLA_GRE_IFLAGS] = { .type = NLA_U16 }, | |
1074 | [IFLA_GRE_OFLAGS] = { .type = NLA_U16 }, | |
1075 | [IFLA_GRE_IKEY] = { .type = NLA_U32 }, | |
1076 | [IFLA_GRE_OKEY] = { .type = NLA_U32 }, | |
4d74f8ba PM |
1077 | [IFLA_GRE_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) }, |
1078 | [IFLA_GRE_REMOTE] = { .len = FIELD_SIZEOF(struct iphdr, daddr) }, | |
c19e654d HX |
1079 | [IFLA_GRE_TTL] = { .type = NLA_U8 }, |
1080 | [IFLA_GRE_TOS] = { .type = NLA_U8 }, | |
1081 | [IFLA_GRE_PMTUDISC] = { .type = NLA_U8 }, | |
4565e991 TH |
1082 | [IFLA_GRE_ENCAP_TYPE] = { .type = NLA_U16 }, |
1083 | [IFLA_GRE_ENCAP_FLAGS] = { .type = NLA_U16 }, | |
1084 | [IFLA_GRE_ENCAP_SPORT] = { .type = NLA_U16 }, | |
1085 | [IFLA_GRE_ENCAP_DPORT] = { .type = NLA_U16 }, | |
2e15ea39 | 1086 | [IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG }, |
c19e654d HX |
1087 | }; |
1088 | ||
1089 | static struct rtnl_link_ops ipgre_link_ops __read_mostly = { | |
1090 | .kind = "gre", | |
1091 | .maxtype = IFLA_GRE_MAX, | |
1092 | .policy = ipgre_policy, | |
1093 | .priv_size = sizeof(struct ip_tunnel), | |
1094 | .setup = ipgre_tunnel_setup, | |
1095 | .validate = ipgre_tunnel_validate, | |
1096 | .newlink = ipgre_newlink, | |
1097 | .changelink = ipgre_changelink, | |
c5441932 | 1098 | .dellink = ip_tunnel_dellink, |
c19e654d HX |
1099 | .get_size = ipgre_get_size, |
1100 | .fill_info = ipgre_fill_info, | |
1728d4fa | 1101 | .get_link_net = ip_tunnel_get_link_net, |
c19e654d HX |
1102 | }; |
1103 | ||
e1a80002 HX |
1104 | static struct rtnl_link_ops ipgre_tap_ops __read_mostly = { |
1105 | .kind = "gretap", | |
1106 | .maxtype = IFLA_GRE_MAX, | |
1107 | .policy = ipgre_policy, | |
1108 | .priv_size = sizeof(struct ip_tunnel), | |
1109 | .setup = ipgre_tap_setup, | |
1110 | .validate = ipgre_tap_validate, | |
1111 | .newlink = ipgre_newlink, | |
1112 | .changelink = ipgre_changelink, | |
c5441932 | 1113 | .dellink = ip_tunnel_dellink, |
e1a80002 HX |
1114 | .get_size = ipgre_get_size, |
1115 | .fill_info = ipgre_fill_info, | |
1728d4fa | 1116 | .get_link_net = ip_tunnel_get_link_net, |
e1a80002 HX |
1117 | }; |
1118 | ||
b2acd1dc PS |
1119 | struct net_device *gretap_fb_dev_create(struct net *net, const char *name, |
1120 | u8 name_assign_type) | |
1121 | { | |
1122 | struct nlattr *tb[IFLA_MAX + 1]; | |
1123 | struct net_device *dev; | |
1124 | struct ip_tunnel *t; | |
1125 | int err; | |
1126 | ||
1127 | memset(&tb, 0, sizeof(tb)); | |
1128 | ||
1129 | dev = rtnl_create_link(net, name, name_assign_type, | |
1130 | &ipgre_tap_ops, tb); | |
1131 | if (IS_ERR(dev)) | |
1132 | return dev; | |
1133 | ||
1134 | /* Configure flow based GRE device. */ | |
1135 | t = netdev_priv(dev); | |
1136 | t->collect_md = true; | |
1137 | ||
1138 | err = ipgre_newlink(net, dev, tb, NULL); | |
1139 | if (err < 0) | |
1140 | goto out; | |
7e059158 DW |
1141 | |
1142 | /* openvswitch users expect packet sizes to be unrestricted, | |
1143 | * so set the largest MTU we can. | |
1144 | */ | |
1145 | err = __ip_tunnel_change_mtu(dev, IP_MAX_MTU, false); | |
1146 | if (err) | |
1147 | goto out; | |
1148 | ||
b2acd1dc PS |
1149 | return dev; |
1150 | out: | |
1151 | free_netdev(dev); | |
1152 | return ERR_PTR(err); | |
1153 | } | |
1154 | EXPORT_SYMBOL_GPL(gretap_fb_dev_create); | |
1155 | ||
c5441932 PS |
1156 | static int __net_init ipgre_tap_init_net(struct net *net) |
1157 | { | |
2e15ea39 | 1158 | return ip_tunnel_init_net(net, gre_tap_net_id, &ipgre_tap_ops, "gretap0"); |
c5441932 PS |
1159 | } |
1160 | ||
1161 | static void __net_exit ipgre_tap_exit_net(struct net *net) | |
1162 | { | |
1163 | struct ip_tunnel_net *itn = net_generic(net, gre_tap_net_id); | |
6c742e71 | 1164 | ip_tunnel_delete_net(itn, &ipgre_tap_ops); |
c5441932 PS |
1165 | } |
1166 | ||
1167 | static struct pernet_operations ipgre_tap_net_ops = { | |
1168 | .init = ipgre_tap_init_net, | |
1169 | .exit = ipgre_tap_exit_net, | |
1170 | .id = &gre_tap_net_id, | |
1171 | .size = sizeof(struct ip_tunnel_net), | |
1172 | }; | |
1da177e4 LT |
1173 | |
1174 | static int __init ipgre_init(void) | |
1175 | { | |
1176 | int err; | |
1177 | ||
058bd4d2 | 1178 | pr_info("GRE over IPv4 tunneling driver\n"); |
1da177e4 | 1179 | |
cfb8fbf2 | 1180 | err = register_pernet_device(&ipgre_net_ops); |
59a4c759 | 1181 | if (err < 0) |
c2892f02 AD |
1182 | return err; |
1183 | ||
c5441932 PS |
1184 | err = register_pernet_device(&ipgre_tap_net_ops); |
1185 | if (err < 0) | |
1186 | goto pnet_tap_faied; | |
1187 | ||
9f57c67c | 1188 | err = gre_add_protocol(&ipgre_protocol, GREPROTO_CISCO); |
c2892f02 | 1189 | if (err < 0) { |
058bd4d2 | 1190 | pr_info("%s: can't add protocol\n", __func__); |
c2892f02 AD |
1191 | goto add_proto_failed; |
1192 | } | |
7daa0004 | 1193 | |
c19e654d HX |
1194 | err = rtnl_link_register(&ipgre_link_ops); |
1195 | if (err < 0) | |
1196 | goto rtnl_link_failed; | |
1197 | ||
e1a80002 HX |
1198 | err = rtnl_link_register(&ipgre_tap_ops); |
1199 | if (err < 0) | |
1200 | goto tap_ops_failed; | |
1201 | ||
c5441932 | 1202 | return 0; |
c19e654d | 1203 | |
e1a80002 HX |
1204 | tap_ops_failed: |
1205 | rtnl_link_unregister(&ipgre_link_ops); | |
c19e654d | 1206 | rtnl_link_failed: |
9f57c67c | 1207 | gre_del_protocol(&ipgre_protocol, GREPROTO_CISCO); |
c2892f02 | 1208 | add_proto_failed: |
c5441932 PS |
1209 | unregister_pernet_device(&ipgre_tap_net_ops); |
1210 | pnet_tap_faied: | |
c2892f02 | 1211 | unregister_pernet_device(&ipgre_net_ops); |
c5441932 | 1212 | return err; |
1da177e4 LT |
1213 | } |
1214 | ||
db44575f | 1215 | static void __exit ipgre_fini(void) |
1da177e4 | 1216 | { |
e1a80002 | 1217 | rtnl_link_unregister(&ipgre_tap_ops); |
c19e654d | 1218 | rtnl_link_unregister(&ipgre_link_ops); |
9f57c67c | 1219 | gre_del_protocol(&ipgre_protocol, GREPROTO_CISCO); |
c5441932 | 1220 | unregister_pernet_device(&ipgre_tap_net_ops); |
c2892f02 | 1221 | unregister_pernet_device(&ipgre_net_ops); |
1da177e4 LT |
1222 | } |
1223 | ||
1224 | module_init(ipgre_init); | |
1225 | module_exit(ipgre_fini); | |
1226 | MODULE_LICENSE("GPL"); | |
4d74f8ba PM |
1227 | MODULE_ALIAS_RTNL_LINK("gre"); |
1228 | MODULE_ALIAS_RTNL_LINK("gretap"); | |
8909c9ad | 1229 | MODULE_ALIAS_NETDEV("gre0"); |
c5441932 | 1230 | MODULE_ALIAS_NETDEV("gretap0"); |