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 | |
1da177e4 LT |
52 | /* |
53 | Problems & solutions | |
54 | -------------------- | |
55 | ||
56 | 1. The most important issue is detecting local dead loops. | |
57 | They would cause complete host lockup in transmit, which | |
58 | would be "resolved" by stack overflow or, if queueing is enabled, | |
59 | with infinite looping in net_bh. | |
60 | ||
61 | We cannot track such dead loops during route installation, | |
62 | it is infeasible task. The most general solutions would be | |
63 | to keep skb->encapsulation counter (sort of local ttl), | |
6d0722a2 | 64 | and silently drop packet when it expires. It is a good |
bff52857 | 65 | solution, but it supposes maintaining new variable in ALL |
1da177e4 LT |
66 | skb, even if no tunneling is used. |
67 | ||
6d0722a2 ED |
68 | Current solution: xmit_recursion breaks dead loops. This is a percpu |
69 | counter, since when we enter the first ndo_xmit(), cpu migration is | |
70 | forbidden. We force an exit if this counter reaches RECURSION_LIMIT | |
1da177e4 LT |
71 | |
72 | 2. Networking dead loops would not kill routers, but would really | |
73 | kill network. IP hop limit plays role of "t->recursion" in this case, | |
74 | if we copy it from packet being encapsulated to upper header. | |
75 | It is very good solution, but it introduces two problems: | |
76 | ||
77 | - Routing protocols, using packets with ttl=1 (OSPF, RIP2), | |
78 | do not work over tunnels. | |
79 | - traceroute does not work. I planned to relay ICMP from tunnel, | |
80 | so that this problem would be solved and traceroute output | |
81 | would even more informative. This idea appeared to be wrong: | |
82 | only Linux complies to rfc1812 now (yes, guys, Linux is the only | |
83 | true router now :-)), all routers (at least, in neighbourhood of mine) | |
84 | return only 8 bytes of payload. It is the end. | |
85 | ||
86 | Hence, if we want that OSPF worked or traceroute said something reasonable, | |
87 | we should search for another solution. | |
88 | ||
89 | One of them is to parse packet trying to detect inner encapsulation | |
90 | made by our node. It is difficult or even impossible, especially, | |
bff52857 | 91 | taking into account fragmentation. TO be short, ttl is not solution at all. |
1da177e4 LT |
92 | |
93 | Current solution: The solution was UNEXPECTEDLY SIMPLE. | |
94 | We force DF flag on tunnels with preconfigured hop limit, | |
95 | that is ALL. :-) Well, it does not remove the problem completely, | |
96 | but exponential growth of network traffic is changed to linear | |
97 | (branches, that exceed pmtu are pruned) and tunnel mtu | |
bff52857 | 98 | rapidly degrades to value <68, where looping stops. |
1da177e4 LT |
99 | Yes, it is not good if there exists a router in the loop, |
100 | which does not force DF, even when encapsulating packets have DF set. | |
101 | But it is not our problem! Nobody could accuse us, we made | |
102 | all that we could make. Even if it is your gated who injected | |
103 | fatal route to network, even if it were you who configured | |
104 | fatal static route: you are innocent. :-) | |
105 | ||
1da177e4 LT |
106 | Alexey Kuznetsov. |
107 | */ | |
108 | ||
eccc1bb8 | 109 | static bool log_ecn_error = true; |
110 | module_param(log_ecn_error, bool, 0644); | |
111 | MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN"); | |
112 | ||
c19e654d | 113 | static struct rtnl_link_ops ipgre_link_ops __read_mostly; |
1da177e4 | 114 | static int ipgre_tunnel_init(struct net_device *dev); |
eb8ce741 | 115 | |
f99189b1 | 116 | static int ipgre_net_id __read_mostly; |
c5441932 | 117 | static int gre_tap_net_id __read_mostly; |
1da177e4 | 118 | |
9f57c67c PS |
119 | static void ipgre_err(struct sk_buff *skb, u32 info, |
120 | const struct tnl_ptk_info *tpi) | |
1da177e4 | 121 | { |
1da177e4 | 122 | |
c5441932 PS |
123 | /* All the routers (except for Linux) return only |
124 | 8 bytes of packet payload. It means, that precise relaying of | |
125 | ICMP in the real Internet is absolutely infeasible. | |
1da177e4 | 126 | |
c5441932 PS |
127 | Moreover, Cisco "wise men" put GRE key to the third word |
128 | in GRE header. It makes impossible maintaining even soft | |
129 | state for keyed GRE tunnels with enabled checksum. Tell | |
130 | them "thank you". | |
1da177e4 | 131 | |
c5441932 PS |
132 | Well, I wonder, rfc1812 was written by Cisco employee, |
133 | what the hell these idiots break standards established | |
134 | by themselves??? | |
135 | */ | |
136 | struct net *net = dev_net(skb->dev); | |
137 | struct ip_tunnel_net *itn; | |
96f5a846 | 138 | const struct iphdr *iph; |
88c7664f ACM |
139 | const int type = icmp_hdr(skb)->type; |
140 | const int code = icmp_hdr(skb)->code; | |
20e1954f | 141 | unsigned int data_len = 0; |
1da177e4 | 142 | struct ip_tunnel *t; |
1da177e4 | 143 | |
1da177e4 LT |
144 | switch (type) { |
145 | default: | |
146 | case ICMP_PARAMETERPROB: | |
9f57c67c | 147 | return; |
1da177e4 LT |
148 | |
149 | case ICMP_DEST_UNREACH: | |
150 | switch (code) { | |
151 | case ICMP_SR_FAILED: | |
152 | case ICMP_PORT_UNREACH: | |
153 | /* Impossible event. */ | |
9f57c67c | 154 | return; |
1da177e4 LT |
155 | default: |
156 | /* All others are translated to HOST_UNREACH. | |
157 | rfc2003 contains "deep thoughts" about NET_UNREACH, | |
158 | I believe they are just ether pollution. --ANK | |
159 | */ | |
160 | break; | |
161 | } | |
162 | break; | |
9f57c67c | 163 | |
1da177e4 LT |
164 | case ICMP_TIME_EXCEEDED: |
165 | if (code != ICMP_EXC_TTL) | |
9f57c67c | 166 | return; |
20e1954f | 167 | data_len = icmp_hdr(skb)->un.reserved[1] * 4; /* RFC 4884 4.1 */ |
1da177e4 | 168 | break; |
55be7a9c DM |
169 | |
170 | case ICMP_REDIRECT: | |
171 | break; | |
1da177e4 LT |
172 | } |
173 | ||
bda7bb46 | 174 | if (tpi->proto == htons(ETH_P_TEB)) |
c5441932 PS |
175 | itn = net_generic(net, gre_tap_net_id); |
176 | else | |
177 | itn = net_generic(net, ipgre_net_id); | |
178 | ||
c0c0c50f | 179 | iph = (const struct iphdr *)(icmp_hdr(skb) + 1); |
bda7bb46 PS |
180 | t = ip_tunnel_lookup(itn, skb->dev->ifindex, tpi->flags, |
181 | iph->daddr, iph->saddr, tpi->key); | |
d2083287 | 182 | |
51456b29 | 183 | if (!t) |
9f57c67c | 184 | return; |
36393395 | 185 | |
9b8c6d7b ED |
186 | #if IS_ENABLED(CONFIG_IPV6) |
187 | if (tpi->proto == htons(ETH_P_IPV6) && | |
20e1954f ED |
188 | !ip6_err_gen_icmpv6_unreach(skb, iph->ihl * 4 + tpi->hdr_len, |
189 | type, data_len)) | |
9b8c6d7b ED |
190 | return; |
191 | #endif | |
192 | ||
36393395 | 193 | if (t->parms.iph.daddr == 0 || |
f97c1e0c | 194 | ipv4_is_multicast(t->parms.iph.daddr)) |
9f57c67c | 195 | return; |
1da177e4 LT |
196 | |
197 | if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED) | |
9f57c67c | 198 | return; |
1da177e4 | 199 | |
da6185d8 | 200 | if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO)) |
1da177e4 LT |
201 | t->err_count++; |
202 | else | |
203 | t->err_count = 1; | |
204 | t->err_time = jiffies; | |
9f57c67c PS |
205 | } |
206 | ||
207 | static void gre_err(struct sk_buff *skb, u32 info) | |
208 | { | |
209 | /* All the routers (except for Linux) return only | |
210 | * 8 bytes of packet payload. It means, that precise relaying of | |
211 | * ICMP in the real Internet is absolutely infeasible. | |
212 | * | |
213 | * Moreover, Cisco "wise men" put GRE key to the third word | |
214 | * in GRE header. It makes impossible maintaining even soft | |
215 | * state for keyed | |
216 | * GRE tunnels with enabled checksum. Tell them "thank you". | |
217 | * | |
218 | * Well, I wonder, rfc1812 was written by Cisco employee, | |
219 | * what the hell these idiots break standards established | |
220 | * by themselves??? | |
221 | */ | |
222 | ||
e582615a | 223 | const struct iphdr *iph = (struct iphdr *)skb->data; |
9f57c67c PS |
224 | const int type = icmp_hdr(skb)->type; |
225 | const int code = icmp_hdr(skb)->code; | |
226 | struct tnl_ptk_info tpi; | |
227 | bool csum_err = false; | |
228 | ||
e582615a ED |
229 | if (gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IP), |
230 | iph->ihl * 4) < 0) { | |
9f57c67c PS |
231 | if (!csum_err) /* ignore csum errors. */ |
232 | return; | |
233 | } | |
234 | ||
235 | if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) { | |
236 | ipv4_update_pmtu(skb, dev_net(skb->dev), info, | |
237 | skb->dev->ifindex, 0, IPPROTO_GRE, 0); | |
238 | return; | |
239 | } | |
240 | if (type == ICMP_REDIRECT) { | |
241 | ipv4_redirect(skb, dev_net(skb->dev), skb->dev->ifindex, 0, | |
242 | IPPROTO_GRE, 0); | |
243 | return; | |
244 | } | |
245 | ||
246 | ipgre_err(skb, info, &tpi); | |
1da177e4 LT |
247 | } |
248 | ||
2e15ea39 PS |
249 | static __be64 key_to_tunnel_id(__be32 key) |
250 | { | |
251 | #ifdef __BIG_ENDIAN | |
252 | return (__force __be64)((__force u32)key); | |
253 | #else | |
254 | return (__force __be64)((__force u64)key << 32); | |
255 | #endif | |
256 | } | |
257 | ||
258 | /* Returns the least-significant 32 bits of a __be64. */ | |
259 | static __be32 tunnel_id_to_key(__be64 x) | |
260 | { | |
261 | #ifdef __BIG_ENDIAN | |
262 | return (__force __be32)x; | |
263 | #else | |
264 | return (__force __be32)((__force u64)x >> 32); | |
265 | #endif | |
266 | } | |
267 | ||
125372fa JB |
268 | static int __ipgre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi, |
269 | struct ip_tunnel_net *itn, int hdr_len, bool raw_proto) | |
1da177e4 | 270 | { |
2e15ea39 | 271 | struct metadata_dst *tun_dst = NULL; |
b71d1d42 | 272 | const struct iphdr *iph; |
1da177e4 | 273 | struct ip_tunnel *tunnel; |
1da177e4 | 274 | |
c5441932 | 275 | iph = ip_hdr(skb); |
bda7bb46 PS |
276 | tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, tpi->flags, |
277 | iph->saddr, iph->daddr, tpi->key); | |
e1a80002 | 278 | |
d2083287 | 279 | if (tunnel) { |
125372fa JB |
280 | if (__iptunnel_pull_header(skb, hdr_len, tpi->proto, |
281 | raw_proto, false) < 0) | |
244a797b JB |
282 | goto drop; |
283 | ||
e271c7b4 JB |
284 | if (tunnel->dev->type != ARPHRD_NONE) |
285 | skb_pop_mac_header(skb); | |
286 | else | |
287 | skb_reset_mac_header(skb); | |
2e15ea39 | 288 | if (tunnel->collect_md) { |
c29a70d2 PS |
289 | __be16 flags; |
290 | __be64 tun_id; | |
2e15ea39 | 291 | |
c29a70d2 PS |
292 | flags = tpi->flags & (TUNNEL_CSUM | TUNNEL_KEY); |
293 | tun_id = key_to_tunnel_id(tpi->key); | |
294 | tun_dst = ip_tun_rx_dst(skb, flags, tun_id, 0); | |
2e15ea39 PS |
295 | if (!tun_dst) |
296 | return PACKET_REJECT; | |
2e15ea39 PS |
297 | } |
298 | ||
299 | ip_tunnel_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error); | |
bda7bb46 | 300 | return PACKET_RCVD; |
1da177e4 | 301 | } |
125372fa | 302 | return PACKET_NEXT; |
244a797b JB |
303 | |
304 | drop: | |
305 | kfree_skb(skb); | |
306 | return PACKET_RCVD; | |
1da177e4 LT |
307 | } |
308 | ||
125372fa JB |
309 | static int ipgre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi, |
310 | int hdr_len) | |
311 | { | |
312 | struct net *net = dev_net(skb->dev); | |
313 | struct ip_tunnel_net *itn; | |
314 | int res; | |
315 | ||
316 | if (tpi->proto == htons(ETH_P_TEB)) | |
317 | itn = net_generic(net, gre_tap_net_id); | |
318 | else | |
319 | itn = net_generic(net, ipgre_net_id); | |
320 | ||
321 | res = __ipgre_rcv(skb, tpi, itn, hdr_len, false); | |
322 | if (res == PACKET_NEXT && tpi->proto == htons(ETH_P_TEB)) { | |
323 | /* ipgre tunnels in collect metadata mode should receive | |
324 | * also ETH_P_TEB traffic. | |
325 | */ | |
326 | itn = net_generic(net, ipgre_net_id); | |
327 | res = __ipgre_rcv(skb, tpi, itn, hdr_len, true); | |
328 | } | |
329 | return res; | |
330 | } | |
331 | ||
9f57c67c PS |
332 | static int gre_rcv(struct sk_buff *skb) |
333 | { | |
334 | struct tnl_ptk_info tpi; | |
335 | bool csum_err = false; | |
95f5c64c | 336 | int hdr_len; |
9f57c67c PS |
337 | |
338 | #ifdef CONFIG_NET_IPGRE_BROADCAST | |
339 | if (ipv4_is_multicast(ip_hdr(skb)->daddr)) { | |
340 | /* Looped back packet, drop it! */ | |
341 | if (rt_is_output_route(skb_rtable(skb))) | |
342 | goto drop; | |
343 | } | |
344 | #endif | |
345 | ||
e582615a | 346 | hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IP), 0); |
f132ae7c | 347 | if (hdr_len < 0) |
95f5c64c TH |
348 | goto drop; |
349 | ||
244a797b | 350 | if (ipgre_rcv(skb, &tpi, hdr_len) == PACKET_RCVD) |
9f57c67c PS |
351 | return 0; |
352 | ||
353 | icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0); | |
354 | drop: | |
355 | kfree_skb(skb); | |
356 | return 0; | |
357 | } | |
358 | ||
c5441932 PS |
359 | static void __gre_xmit(struct sk_buff *skb, struct net_device *dev, |
360 | const struct iphdr *tnl_params, | |
361 | __be16 proto) | |
362 | { | |
363 | struct ip_tunnel *tunnel = netdev_priv(dev); | |
1da177e4 | 364 | |
c5441932 PS |
365 | if (tunnel->parms.o_flags & TUNNEL_SEQ) |
366 | tunnel->o_seqno++; | |
1da177e4 | 367 | |
c5441932 | 368 | /* Push GRE header. */ |
182a352d TH |
369 | gre_build_header(skb, tunnel->tun_hlen, |
370 | tunnel->parms.o_flags, proto, tunnel->parms.o_key, | |
371 | htonl(tunnel->o_seqno)); | |
54bc9bac | 372 | |
bf3d6a8f | 373 | ip_tunnel_xmit(skb, dev, tnl_params, tnl_params->protocol); |
c5441932 | 374 | } |
1da177e4 | 375 | |
aed069df | 376 | static int gre_handle_offloads(struct sk_buff *skb, bool csum) |
b2acd1dc | 377 | { |
6fa79666 | 378 | return iptunnel_handle_offloads(skb, csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE); |
b2acd1dc PS |
379 | } |
380 | ||
fc4099f1 PS |
381 | static struct rtable *gre_get_rt(struct sk_buff *skb, |
382 | struct net_device *dev, | |
383 | struct flowi4 *fl, | |
384 | const struct ip_tunnel_key *key) | |
385 | { | |
386 | struct net *net = dev_net(dev); | |
387 | ||
388 | memset(fl, 0, sizeof(*fl)); | |
389 | fl->daddr = key->u.ipv4.dst; | |
390 | fl->saddr = key->u.ipv4.src; | |
391 | fl->flowi4_tos = RT_TOS(key->tos); | |
392 | fl->flowi4_mark = skb->mark; | |
393 | fl->flowi4_proto = IPPROTO_GRE; | |
394 | ||
395 | return ip_route_output_key(net, fl); | |
396 | } | |
397 | ||
2090714e JB |
398 | static void gre_fb_xmit(struct sk_buff *skb, struct net_device *dev, |
399 | __be16 proto) | |
2e15ea39 PS |
400 | { |
401 | struct ip_tunnel_info *tun_info; | |
2e15ea39 | 402 | const struct ip_tunnel_key *key; |
db3c6139 | 403 | struct rtable *rt = NULL; |
2e15ea39 | 404 | struct flowi4 fl; |
2e15ea39 PS |
405 | int min_headroom; |
406 | int tunnel_hlen; | |
407 | __be16 df, flags; | |
db3c6139 | 408 | bool use_cache; |
2e15ea39 PS |
409 | int err; |
410 | ||
61adedf3 | 411 | tun_info = skb_tunnel_info(skb); |
7f9562a1 JB |
412 | if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) || |
413 | ip_tunnel_info_af(tun_info) != AF_INET)) | |
2e15ea39 PS |
414 | goto err_free_skb; |
415 | ||
416 | key = &tun_info->key; | |
db3c6139 DB |
417 | use_cache = ip_tunnel_dst_cache_usable(skb, tun_info); |
418 | if (use_cache) | |
419 | rt = dst_cache_get_ip4(&tun_info->dst_cache, &fl.saddr); | |
3c1cb4d2 PA |
420 | if (!rt) { |
421 | rt = gre_get_rt(skb, dev, &fl, key); | |
422 | if (IS_ERR(rt)) | |
423 | goto err_free_skb; | |
db3c6139 | 424 | if (use_cache) |
3c1cb4d2 PA |
425 | dst_cache_set_ip4(&tun_info->dst_cache, &rt->dst, |
426 | fl.saddr); | |
427 | } | |
2e15ea39 | 428 | |
95f5c64c | 429 | tunnel_hlen = gre_calc_hlen(key->tun_flags); |
2e15ea39 PS |
430 | |
431 | min_headroom = LL_RESERVED_SPACE(rt->dst.dev) + rt->dst.header_len | |
432 | + tunnel_hlen + sizeof(struct iphdr); | |
433 | if (skb_headroom(skb) < min_headroom || skb_header_cloned(skb)) { | |
434 | int head_delta = SKB_DATA_ALIGN(min_headroom - | |
435 | skb_headroom(skb) + | |
436 | 16); | |
437 | err = pskb_expand_head(skb, max_t(int, head_delta, 0), | |
438 | 0, GFP_ATOMIC); | |
439 | if (unlikely(err)) | |
440 | goto err_free_rt; | |
441 | } | |
442 | ||
443 | /* Push Tunnel header. */ | |
aed069df | 444 | if (gre_handle_offloads(skb, !!(tun_info->key.tun_flags & TUNNEL_CSUM))) |
2e15ea39 | 445 | goto err_free_rt; |
2e15ea39 PS |
446 | |
447 | flags = tun_info->key.tun_flags & (TUNNEL_CSUM | TUNNEL_KEY); | |
cba65321 | 448 | gre_build_header(skb, tunnel_hlen, flags, proto, |
182a352d | 449 | tunnel_id_to_key(tun_info->key.tun_id), 0); |
2e15ea39 PS |
450 | |
451 | df = key->tun_flags & TUNNEL_DONT_FRAGMENT ? htons(IP_DF) : 0; | |
039f5062 PS |
452 | |
453 | iptunnel_xmit(skb->sk, rt, skb, fl.saddr, key->u.ipv4.dst, IPPROTO_GRE, | |
454 | key->tos, key->ttl, df, false); | |
2e15ea39 PS |
455 | return; |
456 | ||
457 | err_free_rt: | |
458 | ip_rt_put(rt); | |
459 | err_free_skb: | |
460 | kfree_skb(skb); | |
461 | dev->stats.tx_dropped++; | |
462 | } | |
463 | ||
fc4099f1 PS |
464 | static int gre_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb) |
465 | { | |
466 | struct ip_tunnel_info *info = skb_tunnel_info(skb); | |
467 | struct rtable *rt; | |
468 | struct flowi4 fl4; | |
469 | ||
470 | if (ip_tunnel_info_af(info) != AF_INET) | |
471 | return -EINVAL; | |
472 | ||
473 | rt = gre_get_rt(skb, dev, &fl4, &info->key); | |
474 | if (IS_ERR(rt)) | |
475 | return PTR_ERR(rt); | |
476 | ||
477 | ip_rt_put(rt); | |
478 | info->key.u.ipv4.src = fl4.saddr; | |
479 | return 0; | |
480 | } | |
481 | ||
c5441932 PS |
482 | static netdev_tx_t ipgre_xmit(struct sk_buff *skb, |
483 | struct net_device *dev) | |
484 | { | |
485 | struct ip_tunnel *tunnel = netdev_priv(dev); | |
486 | const struct iphdr *tnl_params; | |
1da177e4 | 487 | |
2e15ea39 | 488 | if (tunnel->collect_md) { |
2090714e | 489 | gre_fb_xmit(skb, dev, skb->protocol); |
2e15ea39 PS |
490 | return NETDEV_TX_OK; |
491 | } | |
492 | ||
c5441932 PS |
493 | if (dev->header_ops) { |
494 | /* Need space for new headers */ | |
495 | if (skb_cow_head(skb, dev->needed_headroom - | |
2bac7cb3 | 496 | (tunnel->hlen + sizeof(struct iphdr)))) |
c5441932 | 497 | goto free_skb; |
1da177e4 | 498 | |
c5441932 | 499 | tnl_params = (const struct iphdr *)skb->data; |
1da177e4 | 500 | |
c5441932 PS |
501 | /* Pull skb since ip_tunnel_xmit() needs skb->data pointing |
502 | * to gre header. | |
503 | */ | |
504 | skb_pull(skb, tunnel->hlen + sizeof(struct iphdr)); | |
8a0033a9 | 505 | skb_reset_mac_header(skb); |
c5441932 PS |
506 | } else { |
507 | if (skb_cow_head(skb, dev->needed_headroom)) | |
508 | goto free_skb; | |
1da177e4 | 509 | |
c5441932 | 510 | tnl_params = &tunnel->parms.iph; |
1da177e4 LT |
511 | } |
512 | ||
aed069df AD |
513 | if (gre_handle_offloads(skb, !!(tunnel->parms.o_flags & TUNNEL_CSUM))) |
514 | goto free_skb; | |
8a0033a9 | 515 | |
c5441932 | 516 | __gre_xmit(skb, dev, tnl_params, skb->protocol); |
6ed10654 | 517 | return NETDEV_TX_OK; |
1da177e4 | 518 | |
c5441932 | 519 | free_skb: |
3acfa1e7 | 520 | kfree_skb(skb); |
c5441932 | 521 | dev->stats.tx_dropped++; |
6ed10654 | 522 | return NETDEV_TX_OK; |
1da177e4 LT |
523 | } |
524 | ||
c5441932 PS |
525 | static netdev_tx_t gre_tap_xmit(struct sk_buff *skb, |
526 | struct net_device *dev) | |
ee34c1eb | 527 | { |
c5441932 | 528 | struct ip_tunnel *tunnel = netdev_priv(dev); |
ee34c1eb | 529 | |
2e15ea39 | 530 | if (tunnel->collect_md) { |
2090714e | 531 | gre_fb_xmit(skb, dev, htons(ETH_P_TEB)); |
2e15ea39 PS |
532 | return NETDEV_TX_OK; |
533 | } | |
534 | ||
aed069df AD |
535 | if (gre_handle_offloads(skb, !!(tunnel->parms.o_flags & TUNNEL_CSUM))) |
536 | goto free_skb; | |
ee34c1eb | 537 | |
c5441932 PS |
538 | if (skb_cow_head(skb, dev->needed_headroom)) |
539 | goto free_skb; | |
42aa9162 | 540 | |
c5441932 | 541 | __gre_xmit(skb, dev, &tunnel->parms.iph, htons(ETH_P_TEB)); |
c5441932 | 542 | return NETDEV_TX_OK; |
ee34c1eb | 543 | |
c5441932 | 544 | free_skb: |
3acfa1e7 | 545 | kfree_skb(skb); |
c5441932 PS |
546 | dev->stats.tx_dropped++; |
547 | return NETDEV_TX_OK; | |
ee34c1eb MS |
548 | } |
549 | ||
c5441932 PS |
550 | static int ipgre_tunnel_ioctl(struct net_device *dev, |
551 | struct ifreq *ifr, int cmd) | |
1da177e4 | 552 | { |
4565e991 | 553 | int err; |
1da177e4 | 554 | struct ip_tunnel_parm p; |
1da177e4 | 555 | |
c5441932 PS |
556 | if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) |
557 | return -EFAULT; | |
6c734fb8 CW |
558 | if (cmd == SIOCADDTUNNEL || cmd == SIOCCHGTUNNEL) { |
559 | if (p.iph.version != 4 || p.iph.protocol != IPPROTO_GRE || | |
560 | p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)) || | |
561 | ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))) | |
562 | return -EINVAL; | |
1da177e4 | 563 | } |
c5441932 PS |
564 | p.i_flags = gre_flags_to_tnl_flags(p.i_flags); |
565 | p.o_flags = gre_flags_to_tnl_flags(p.o_flags); | |
1da177e4 | 566 | |
c5441932 PS |
567 | err = ip_tunnel_ioctl(dev, &p, cmd); |
568 | if (err) | |
569 | return err; | |
1da177e4 | 570 | |
95f5c64c TH |
571 | p.i_flags = gre_tnl_flags_to_gre_flags(p.i_flags); |
572 | p.o_flags = gre_tnl_flags_to_gre_flags(p.o_flags); | |
c5441932 PS |
573 | |
574 | if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) | |
575 | return -EFAULT; | |
1da177e4 LT |
576 | return 0; |
577 | } | |
578 | ||
1da177e4 LT |
579 | /* Nice toy. Unfortunately, useless in real life :-) |
580 | It allows to construct virtual multiprotocol broadcast "LAN" | |
581 | over the Internet, provided multicast routing is tuned. | |
582 | ||
583 | ||
584 | I have no idea was this bicycle invented before me, | |
585 | so that I had to set ARPHRD_IPGRE to a random value. | |
586 | I have an impression, that Cisco could make something similar, | |
587 | but this feature is apparently missing in IOS<=11.2(8). | |
e905a9ed | 588 | |
1da177e4 LT |
589 | I set up 10.66.66/24 and fec0:6666:6666::0/96 as virtual networks |
590 | with broadcast 224.66.66.66. If you have access to mbone, play with me :-) | |
591 | ||
592 | ping -t 255 224.66.66.66 | |
593 | ||
594 | If nobody answers, mbone does not work. | |
595 | ||
596 | ip tunnel add Universe mode gre remote 224.66.66.66 local <Your_real_addr> ttl 255 | |
597 | ip addr add 10.66.66.<somewhat>/24 dev Universe | |
598 | ifconfig Universe up | |
599 | ifconfig Universe add fe80::<Your_real_addr>/10 | |
600 | ifconfig Universe add fec0:6666:6666::<Your_real_addr>/96 | |
601 | ftp 10.66.66.66 | |
602 | ... | |
603 | ftp fec0:6666:6666::193.233.7.65 | |
604 | ... | |
1da177e4 | 605 | */ |
3b04ddde SH |
606 | static int ipgre_header(struct sk_buff *skb, struct net_device *dev, |
607 | unsigned short type, | |
1507850b | 608 | const void *daddr, const void *saddr, unsigned int len) |
1da177e4 | 609 | { |
2941a486 | 610 | struct ip_tunnel *t = netdev_priv(dev); |
c5441932 PS |
611 | struct iphdr *iph; |
612 | struct gre_base_hdr *greh; | |
1da177e4 | 613 | |
c5441932 PS |
614 | iph = (struct iphdr *)skb_push(skb, t->hlen + sizeof(*iph)); |
615 | greh = (struct gre_base_hdr *)(iph+1); | |
95f5c64c | 616 | greh->flags = gre_tnl_flags_to_gre_flags(t->parms.o_flags); |
c5441932 | 617 | greh->protocol = htons(type); |
1da177e4 | 618 | |
c5441932 | 619 | memcpy(iph, &t->parms.iph, sizeof(struct iphdr)); |
e905a9ed | 620 | |
c5441932 | 621 | /* Set the source hardware address. */ |
1da177e4 LT |
622 | if (saddr) |
623 | memcpy(&iph->saddr, saddr, 4); | |
6d55cb91 | 624 | if (daddr) |
1da177e4 | 625 | memcpy(&iph->daddr, daddr, 4); |
6d55cb91 | 626 | if (iph->daddr) |
77a482bd | 627 | return t->hlen + sizeof(*iph); |
e905a9ed | 628 | |
c5441932 | 629 | return -(t->hlen + sizeof(*iph)); |
1da177e4 LT |
630 | } |
631 | ||
6a5f44d7 TT |
632 | static int ipgre_header_parse(const struct sk_buff *skb, unsigned char *haddr) |
633 | { | |
b71d1d42 | 634 | const struct iphdr *iph = (const struct iphdr *) skb_mac_header(skb); |
6a5f44d7 TT |
635 | memcpy(haddr, &iph->saddr, 4); |
636 | return 4; | |
637 | } | |
638 | ||
3b04ddde SH |
639 | static const struct header_ops ipgre_header_ops = { |
640 | .create = ipgre_header, | |
6a5f44d7 | 641 | .parse = ipgre_header_parse, |
3b04ddde SH |
642 | }; |
643 | ||
6a5f44d7 | 644 | #ifdef CONFIG_NET_IPGRE_BROADCAST |
1da177e4 LT |
645 | static int ipgre_open(struct net_device *dev) |
646 | { | |
2941a486 | 647 | struct ip_tunnel *t = netdev_priv(dev); |
1da177e4 | 648 | |
f97c1e0c | 649 | if (ipv4_is_multicast(t->parms.iph.daddr)) { |
cbb1e85f DM |
650 | struct flowi4 fl4; |
651 | struct rtable *rt; | |
652 | ||
b57708ad | 653 | rt = ip_route_output_gre(t->net, &fl4, |
cbb1e85f DM |
654 | t->parms.iph.daddr, |
655 | t->parms.iph.saddr, | |
656 | t->parms.o_key, | |
657 | RT_TOS(t->parms.iph.tos), | |
658 | t->parms.link); | |
b23dd4fe | 659 | if (IS_ERR(rt)) |
1da177e4 | 660 | return -EADDRNOTAVAIL; |
d8d1f30b | 661 | dev = rt->dst.dev; |
1da177e4 | 662 | ip_rt_put(rt); |
51456b29 | 663 | if (!__in_dev_get_rtnl(dev)) |
1da177e4 LT |
664 | return -EADDRNOTAVAIL; |
665 | t->mlink = dev->ifindex; | |
e5ed6399 | 666 | ip_mc_inc_group(__in_dev_get_rtnl(dev), t->parms.iph.daddr); |
1da177e4 LT |
667 | } |
668 | return 0; | |
669 | } | |
670 | ||
671 | static int ipgre_close(struct net_device *dev) | |
672 | { | |
2941a486 | 673 | struct ip_tunnel *t = netdev_priv(dev); |
b8c26a33 | 674 | |
f97c1e0c | 675 | if (ipv4_is_multicast(t->parms.iph.daddr) && t->mlink) { |
7fee0ca2 | 676 | struct in_device *in_dev; |
b57708ad | 677 | in_dev = inetdev_by_index(t->net, t->mlink); |
8723e1b4 | 678 | if (in_dev) |
1da177e4 | 679 | ip_mc_dec_group(in_dev, t->parms.iph.daddr); |
1da177e4 LT |
680 | } |
681 | return 0; | |
682 | } | |
1da177e4 LT |
683 | #endif |
684 | ||
b8c26a33 SH |
685 | static const struct net_device_ops ipgre_netdev_ops = { |
686 | .ndo_init = ipgre_tunnel_init, | |
c5441932 | 687 | .ndo_uninit = ip_tunnel_uninit, |
b8c26a33 SH |
688 | #ifdef CONFIG_NET_IPGRE_BROADCAST |
689 | .ndo_open = ipgre_open, | |
690 | .ndo_stop = ipgre_close, | |
691 | #endif | |
c5441932 | 692 | .ndo_start_xmit = ipgre_xmit, |
b8c26a33 | 693 | .ndo_do_ioctl = ipgre_tunnel_ioctl, |
c5441932 PS |
694 | .ndo_change_mtu = ip_tunnel_change_mtu, |
695 | .ndo_get_stats64 = ip_tunnel_get_stats64, | |
1e99584b | 696 | .ndo_get_iflink = ip_tunnel_get_iflink, |
b8c26a33 SH |
697 | }; |
698 | ||
6b78f16e ED |
699 | #define GRE_FEATURES (NETIF_F_SG | \ |
700 | NETIF_F_FRAGLIST | \ | |
701 | NETIF_F_HIGHDMA | \ | |
702 | NETIF_F_HW_CSUM) | |
703 | ||
1da177e4 LT |
704 | static void ipgre_tunnel_setup(struct net_device *dev) |
705 | { | |
b8c26a33 | 706 | dev->netdev_ops = &ipgre_netdev_ops; |
5a455275 | 707 | dev->type = ARPHRD_IPGRE; |
c5441932 PS |
708 | ip_tunnel_setup(dev, ipgre_net_id); |
709 | } | |
1da177e4 | 710 | |
c5441932 PS |
711 | static void __gre_tunnel_init(struct net_device *dev) |
712 | { | |
713 | struct ip_tunnel *tunnel; | |
4565e991 | 714 | int t_hlen; |
c5441932 PS |
715 | |
716 | tunnel = netdev_priv(dev); | |
95f5c64c | 717 | tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags); |
c5441932 PS |
718 | tunnel->parms.iph.protocol = IPPROTO_GRE; |
719 | ||
4565e991 TH |
720 | tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen; |
721 | ||
722 | t_hlen = tunnel->hlen + sizeof(struct iphdr); | |
723 | ||
724 | dev->needed_headroom = LL_MAX_HEADER + t_hlen + 4; | |
725 | dev->mtu = ETH_DATA_LEN - t_hlen - 4; | |
6b78f16e | 726 | |
b57708ad | 727 | dev->features |= GRE_FEATURES; |
6b78f16e | 728 | dev->hw_features |= GRE_FEATURES; |
c5441932 PS |
729 | |
730 | if (!(tunnel->parms.o_flags & TUNNEL_SEQ)) { | |
a0ca153f AD |
731 | /* TCP offload with GRE SEQ is not supported, nor |
732 | * can we support 2 levels of outer headers requiring | |
733 | * an update. | |
734 | */ | |
735 | if (!(tunnel->parms.o_flags & TUNNEL_CSUM) || | |
736 | (tunnel->encap.type == TUNNEL_ENCAP_NONE)) { | |
737 | dev->features |= NETIF_F_GSO_SOFTWARE; | |
738 | dev->hw_features |= NETIF_F_GSO_SOFTWARE; | |
739 | } | |
740 | ||
c5441932 PS |
741 | /* Can use a lockless transmit, unless we generate |
742 | * output sequences | |
743 | */ | |
744 | dev->features |= NETIF_F_LLTX; | |
745 | } | |
1da177e4 LT |
746 | } |
747 | ||
748 | static int ipgre_tunnel_init(struct net_device *dev) | |
749 | { | |
c5441932 PS |
750 | struct ip_tunnel *tunnel = netdev_priv(dev); |
751 | struct iphdr *iph = &tunnel->parms.iph; | |
1da177e4 | 752 | |
c5441932 | 753 | __gre_tunnel_init(dev); |
1da177e4 | 754 | |
c5441932 PS |
755 | memcpy(dev->dev_addr, &iph->saddr, 4); |
756 | memcpy(dev->broadcast, &iph->daddr, 4); | |
1da177e4 | 757 | |
c5441932 | 758 | dev->flags = IFF_NOARP; |
02875878 | 759 | netif_keep_dst(dev); |
c5441932 | 760 | dev->addr_len = 4; |
1da177e4 | 761 | |
a64b04d8 | 762 | if (iph->daddr && !tunnel->collect_md) { |
1da177e4 | 763 | #ifdef CONFIG_NET_IPGRE_BROADCAST |
f97c1e0c | 764 | if (ipv4_is_multicast(iph->daddr)) { |
1da177e4 LT |
765 | if (!iph->saddr) |
766 | return -EINVAL; | |
767 | dev->flags = IFF_BROADCAST; | |
3b04ddde | 768 | dev->header_ops = &ipgre_header_ops; |
1da177e4 LT |
769 | } |
770 | #endif | |
a64b04d8 | 771 | } else if (!tunnel->collect_md) { |
6a5f44d7 | 772 | dev->header_ops = &ipgre_header_ops; |
a64b04d8 | 773 | } |
1da177e4 | 774 | |
c5441932 | 775 | return ip_tunnel_init(dev); |
1da177e4 LT |
776 | } |
777 | ||
9f57c67c PS |
778 | static const struct gre_protocol ipgre_protocol = { |
779 | .handler = gre_rcv, | |
780 | .err_handler = gre_err, | |
1da177e4 LT |
781 | }; |
782 | ||
2c8c1e72 | 783 | static int __net_init ipgre_init_net(struct net *net) |
59a4c759 | 784 | { |
c5441932 | 785 | return ip_tunnel_init_net(net, ipgre_net_id, &ipgre_link_ops, NULL); |
59a4c759 PE |
786 | } |
787 | ||
2c8c1e72 | 788 | static void __net_exit ipgre_exit_net(struct net *net) |
59a4c759 | 789 | { |
c5441932 | 790 | struct ip_tunnel_net *itn = net_generic(net, ipgre_net_id); |
6c742e71 | 791 | ip_tunnel_delete_net(itn, &ipgre_link_ops); |
59a4c759 PE |
792 | } |
793 | ||
794 | static struct pernet_operations ipgre_net_ops = { | |
795 | .init = ipgre_init_net, | |
796 | .exit = ipgre_exit_net, | |
cfb8fbf2 | 797 | .id = &ipgre_net_id, |
c5441932 | 798 | .size = sizeof(struct ip_tunnel_net), |
59a4c759 | 799 | }; |
1da177e4 | 800 | |
c19e654d HX |
801 | static int ipgre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[]) |
802 | { | |
803 | __be16 flags; | |
804 | ||
805 | if (!data) | |
806 | return 0; | |
807 | ||
808 | flags = 0; | |
809 | if (data[IFLA_GRE_IFLAGS]) | |
810 | flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]); | |
811 | if (data[IFLA_GRE_OFLAGS]) | |
812 | flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]); | |
813 | if (flags & (GRE_VERSION|GRE_ROUTING)) | |
814 | return -EINVAL; | |
815 | ||
946b636f JB |
816 | if (data[IFLA_GRE_COLLECT_METADATA] && |
817 | data[IFLA_GRE_ENCAP_TYPE] && | |
818 | nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]) != TUNNEL_ENCAP_NONE) | |
819 | return -EINVAL; | |
820 | ||
c19e654d HX |
821 | return 0; |
822 | } | |
823 | ||
e1a80002 HX |
824 | static int ipgre_tap_validate(struct nlattr *tb[], struct nlattr *data[]) |
825 | { | |
826 | __be32 daddr; | |
827 | ||
828 | if (tb[IFLA_ADDRESS]) { | |
829 | if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) | |
830 | return -EINVAL; | |
831 | if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) | |
832 | return -EADDRNOTAVAIL; | |
833 | } | |
834 | ||
835 | if (!data) | |
836 | goto out; | |
837 | ||
838 | if (data[IFLA_GRE_REMOTE]) { | |
839 | memcpy(&daddr, nla_data(data[IFLA_GRE_REMOTE]), 4); | |
840 | if (!daddr) | |
841 | return -EINVAL; | |
842 | } | |
843 | ||
844 | out: | |
845 | return ipgre_tunnel_validate(tb, data); | |
846 | } | |
847 | ||
22a59be8 | 848 | static int ipgre_netlink_parms(struct net_device *dev, |
2e15ea39 PS |
849 | struct nlattr *data[], |
850 | struct nlattr *tb[], | |
851 | struct ip_tunnel_parm *parms) | |
c19e654d | 852 | { |
22a59be8 PP |
853 | struct ip_tunnel *t = netdev_priv(dev); |
854 | ||
7bb82d92 | 855 | memset(parms, 0, sizeof(*parms)); |
c19e654d HX |
856 | |
857 | parms->iph.protocol = IPPROTO_GRE; | |
858 | ||
859 | if (!data) | |
22a59be8 | 860 | return 0; |
c19e654d HX |
861 | |
862 | if (data[IFLA_GRE_LINK]) | |
863 | parms->link = nla_get_u32(data[IFLA_GRE_LINK]); | |
864 | ||
865 | if (data[IFLA_GRE_IFLAGS]) | |
c5441932 | 866 | parms->i_flags = gre_flags_to_tnl_flags(nla_get_be16(data[IFLA_GRE_IFLAGS])); |
c19e654d HX |
867 | |
868 | if (data[IFLA_GRE_OFLAGS]) | |
c5441932 | 869 | parms->o_flags = gre_flags_to_tnl_flags(nla_get_be16(data[IFLA_GRE_OFLAGS])); |
c19e654d HX |
870 | |
871 | if (data[IFLA_GRE_IKEY]) | |
872 | parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]); | |
873 | ||
874 | if (data[IFLA_GRE_OKEY]) | |
875 | parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]); | |
876 | ||
877 | if (data[IFLA_GRE_LOCAL]) | |
67b61f6c | 878 | parms->iph.saddr = nla_get_in_addr(data[IFLA_GRE_LOCAL]); |
c19e654d HX |
879 | |
880 | if (data[IFLA_GRE_REMOTE]) | |
67b61f6c | 881 | parms->iph.daddr = nla_get_in_addr(data[IFLA_GRE_REMOTE]); |
c19e654d HX |
882 | |
883 | if (data[IFLA_GRE_TTL]) | |
884 | parms->iph.ttl = nla_get_u8(data[IFLA_GRE_TTL]); | |
885 | ||
886 | if (data[IFLA_GRE_TOS]) | |
887 | parms->iph.tos = nla_get_u8(data[IFLA_GRE_TOS]); | |
888 | ||
22a59be8 PP |
889 | if (!data[IFLA_GRE_PMTUDISC] || nla_get_u8(data[IFLA_GRE_PMTUDISC])) { |
890 | if (t->ignore_df) | |
891 | return -EINVAL; | |
c19e654d | 892 | parms->iph.frag_off = htons(IP_DF); |
22a59be8 | 893 | } |
2e15ea39 PS |
894 | |
895 | if (data[IFLA_GRE_COLLECT_METADATA]) { | |
2e15ea39 | 896 | t->collect_md = true; |
e271c7b4 JB |
897 | if (dev->type == ARPHRD_IPGRE) |
898 | dev->type = ARPHRD_NONE; | |
2e15ea39 | 899 | } |
22a59be8 PP |
900 | |
901 | if (data[IFLA_GRE_IGNORE_DF]) { | |
902 | if (nla_get_u8(data[IFLA_GRE_IGNORE_DF]) | |
903 | && (parms->iph.frag_off & htons(IP_DF))) | |
904 | return -EINVAL; | |
905 | t->ignore_df = !!nla_get_u8(data[IFLA_GRE_IGNORE_DF]); | |
906 | } | |
907 | ||
908 | return 0; | |
c19e654d HX |
909 | } |
910 | ||
4565e991 TH |
911 | /* This function returns true when ENCAP attributes are present in the nl msg */ |
912 | static bool ipgre_netlink_encap_parms(struct nlattr *data[], | |
913 | struct ip_tunnel_encap *ipencap) | |
914 | { | |
915 | bool ret = false; | |
916 | ||
917 | memset(ipencap, 0, sizeof(*ipencap)); | |
918 | ||
919 | if (!data) | |
920 | return ret; | |
921 | ||
922 | if (data[IFLA_GRE_ENCAP_TYPE]) { | |
923 | ret = true; | |
924 | ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]); | |
925 | } | |
926 | ||
927 | if (data[IFLA_GRE_ENCAP_FLAGS]) { | |
928 | ret = true; | |
929 | ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]); | |
930 | } | |
931 | ||
932 | if (data[IFLA_GRE_ENCAP_SPORT]) { | |
933 | ret = true; | |
3e97fa70 | 934 | ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]); |
4565e991 TH |
935 | } |
936 | ||
937 | if (data[IFLA_GRE_ENCAP_DPORT]) { | |
938 | ret = true; | |
3e97fa70 | 939 | ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]); |
4565e991 TH |
940 | } |
941 | ||
942 | return ret; | |
943 | } | |
944 | ||
c5441932 | 945 | static int gre_tap_init(struct net_device *dev) |
e1a80002 | 946 | { |
c5441932 | 947 | __gre_tunnel_init(dev); |
bec94d43 | 948 | dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; |
e1a80002 | 949 | |
c5441932 | 950 | return ip_tunnel_init(dev); |
e1a80002 HX |
951 | } |
952 | ||
c5441932 PS |
953 | static const struct net_device_ops gre_tap_netdev_ops = { |
954 | .ndo_init = gre_tap_init, | |
955 | .ndo_uninit = ip_tunnel_uninit, | |
956 | .ndo_start_xmit = gre_tap_xmit, | |
b8c26a33 SH |
957 | .ndo_set_mac_address = eth_mac_addr, |
958 | .ndo_validate_addr = eth_validate_addr, | |
c5441932 PS |
959 | .ndo_change_mtu = ip_tunnel_change_mtu, |
960 | .ndo_get_stats64 = ip_tunnel_get_stats64, | |
1e99584b | 961 | .ndo_get_iflink = ip_tunnel_get_iflink, |
fc4099f1 | 962 | .ndo_fill_metadata_dst = gre_fill_metadata_dst, |
b8c26a33 SH |
963 | }; |
964 | ||
e1a80002 HX |
965 | static void ipgre_tap_setup(struct net_device *dev) |
966 | { | |
e1a80002 | 967 | ether_setup(dev); |
d13b161c JB |
968 | dev->netdev_ops = &gre_tap_netdev_ops; |
969 | dev->priv_flags &= ~IFF_TX_SKB_SHARING; | |
970 | dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; | |
c5441932 | 971 | ip_tunnel_setup(dev, gre_tap_net_id); |
e1a80002 HX |
972 | } |
973 | ||
c5441932 PS |
974 | static int ipgre_newlink(struct net *src_net, struct net_device *dev, |
975 | struct nlattr *tb[], struct nlattr *data[]) | |
c19e654d | 976 | { |
c5441932 | 977 | struct ip_tunnel_parm p; |
4565e991 | 978 | struct ip_tunnel_encap ipencap; |
22a59be8 | 979 | int err; |
4565e991 TH |
980 | |
981 | if (ipgre_netlink_encap_parms(data, &ipencap)) { | |
982 | struct ip_tunnel *t = netdev_priv(dev); | |
22a59be8 | 983 | err = ip_tunnel_encap_setup(t, &ipencap); |
4565e991 TH |
984 | |
985 | if (err < 0) | |
986 | return err; | |
987 | } | |
c19e654d | 988 | |
22a59be8 PP |
989 | err = ipgre_netlink_parms(dev, data, tb, &p); |
990 | if (err < 0) | |
991 | return err; | |
c5441932 | 992 | return ip_tunnel_newlink(dev, tb, &p); |
c19e654d HX |
993 | } |
994 | ||
995 | static int ipgre_changelink(struct net_device *dev, struct nlattr *tb[], | |
996 | struct nlattr *data[]) | |
997 | { | |
c19e654d | 998 | struct ip_tunnel_parm p; |
4565e991 | 999 | struct ip_tunnel_encap ipencap; |
22a59be8 | 1000 | int err; |
4565e991 TH |
1001 | |
1002 | if (ipgre_netlink_encap_parms(data, &ipencap)) { | |
1003 | struct ip_tunnel *t = netdev_priv(dev); | |
22a59be8 | 1004 | err = ip_tunnel_encap_setup(t, &ipencap); |
4565e991 TH |
1005 | |
1006 | if (err < 0) | |
1007 | return err; | |
1008 | } | |
c19e654d | 1009 | |
22a59be8 PP |
1010 | err = ipgre_netlink_parms(dev, data, tb, &p); |
1011 | if (err < 0) | |
1012 | return err; | |
c5441932 | 1013 | return ip_tunnel_changelink(dev, tb, &p); |
c19e654d HX |
1014 | } |
1015 | ||
1016 | static size_t ipgre_get_size(const struct net_device *dev) | |
1017 | { | |
1018 | return | |
1019 | /* IFLA_GRE_LINK */ | |
1020 | nla_total_size(4) + | |
1021 | /* IFLA_GRE_IFLAGS */ | |
1022 | nla_total_size(2) + | |
1023 | /* IFLA_GRE_OFLAGS */ | |
1024 | nla_total_size(2) + | |
1025 | /* IFLA_GRE_IKEY */ | |
1026 | nla_total_size(4) + | |
1027 | /* IFLA_GRE_OKEY */ | |
1028 | nla_total_size(4) + | |
1029 | /* IFLA_GRE_LOCAL */ | |
1030 | nla_total_size(4) + | |
1031 | /* IFLA_GRE_REMOTE */ | |
1032 | nla_total_size(4) + | |
1033 | /* IFLA_GRE_TTL */ | |
1034 | nla_total_size(1) + | |
1035 | /* IFLA_GRE_TOS */ | |
1036 | nla_total_size(1) + | |
1037 | /* IFLA_GRE_PMTUDISC */ | |
1038 | nla_total_size(1) + | |
4565e991 TH |
1039 | /* IFLA_GRE_ENCAP_TYPE */ |
1040 | nla_total_size(2) + | |
1041 | /* IFLA_GRE_ENCAP_FLAGS */ | |
1042 | nla_total_size(2) + | |
1043 | /* IFLA_GRE_ENCAP_SPORT */ | |
1044 | nla_total_size(2) + | |
1045 | /* IFLA_GRE_ENCAP_DPORT */ | |
1046 | nla_total_size(2) + | |
2e15ea39 PS |
1047 | /* IFLA_GRE_COLLECT_METADATA */ |
1048 | nla_total_size(0) + | |
22a59be8 PP |
1049 | /* IFLA_GRE_IGNORE_DF */ |
1050 | nla_total_size(1) + | |
c19e654d HX |
1051 | 0; |
1052 | } | |
1053 | ||
1054 | static int ipgre_fill_info(struct sk_buff *skb, const struct net_device *dev) | |
1055 | { | |
1056 | struct ip_tunnel *t = netdev_priv(dev); | |
1057 | struct ip_tunnel_parm *p = &t->parms; | |
1058 | ||
f3756b79 | 1059 | if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) || |
95f5c64c TH |
1060 | nla_put_be16(skb, IFLA_GRE_IFLAGS, |
1061 | gre_tnl_flags_to_gre_flags(p->i_flags)) || | |
1062 | nla_put_be16(skb, IFLA_GRE_OFLAGS, | |
1063 | gre_tnl_flags_to_gre_flags(p->o_flags)) || | |
f3756b79 DM |
1064 | nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) || |
1065 | nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) || | |
930345ea JB |
1066 | nla_put_in_addr(skb, IFLA_GRE_LOCAL, p->iph.saddr) || |
1067 | nla_put_in_addr(skb, IFLA_GRE_REMOTE, p->iph.daddr) || | |
f3756b79 DM |
1068 | nla_put_u8(skb, IFLA_GRE_TTL, p->iph.ttl) || |
1069 | nla_put_u8(skb, IFLA_GRE_TOS, p->iph.tos) || | |
1070 | nla_put_u8(skb, IFLA_GRE_PMTUDISC, | |
1071 | !!(p->iph.frag_off & htons(IP_DF)))) | |
1072 | goto nla_put_failure; | |
4565e991 TH |
1073 | |
1074 | if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE, | |
1075 | t->encap.type) || | |
3e97fa70 SD |
1076 | nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT, |
1077 | t->encap.sport) || | |
1078 | nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT, | |
1079 | t->encap.dport) || | |
4565e991 | 1080 | nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS, |
e1b2cb65 | 1081 | t->encap.flags)) |
4565e991 TH |
1082 | goto nla_put_failure; |
1083 | ||
22a59be8 PP |
1084 | if (nla_put_u8(skb, IFLA_GRE_IGNORE_DF, t->ignore_df)) |
1085 | goto nla_put_failure; | |
1086 | ||
2e15ea39 PS |
1087 | if (t->collect_md) { |
1088 | if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA)) | |
1089 | goto nla_put_failure; | |
1090 | } | |
1091 | ||
c19e654d HX |
1092 | return 0; |
1093 | ||
1094 | nla_put_failure: | |
1095 | return -EMSGSIZE; | |
1096 | } | |
1097 | ||
1098 | static const struct nla_policy ipgre_policy[IFLA_GRE_MAX + 1] = { | |
1099 | [IFLA_GRE_LINK] = { .type = NLA_U32 }, | |
1100 | [IFLA_GRE_IFLAGS] = { .type = NLA_U16 }, | |
1101 | [IFLA_GRE_OFLAGS] = { .type = NLA_U16 }, | |
1102 | [IFLA_GRE_IKEY] = { .type = NLA_U32 }, | |
1103 | [IFLA_GRE_OKEY] = { .type = NLA_U32 }, | |
4d74f8ba PM |
1104 | [IFLA_GRE_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) }, |
1105 | [IFLA_GRE_REMOTE] = { .len = FIELD_SIZEOF(struct iphdr, daddr) }, | |
c19e654d HX |
1106 | [IFLA_GRE_TTL] = { .type = NLA_U8 }, |
1107 | [IFLA_GRE_TOS] = { .type = NLA_U8 }, | |
1108 | [IFLA_GRE_PMTUDISC] = { .type = NLA_U8 }, | |
4565e991 TH |
1109 | [IFLA_GRE_ENCAP_TYPE] = { .type = NLA_U16 }, |
1110 | [IFLA_GRE_ENCAP_FLAGS] = { .type = NLA_U16 }, | |
1111 | [IFLA_GRE_ENCAP_SPORT] = { .type = NLA_U16 }, | |
1112 | [IFLA_GRE_ENCAP_DPORT] = { .type = NLA_U16 }, | |
2e15ea39 | 1113 | [IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG }, |
22a59be8 | 1114 | [IFLA_GRE_IGNORE_DF] = { .type = NLA_U8 }, |
c19e654d HX |
1115 | }; |
1116 | ||
1117 | static struct rtnl_link_ops ipgre_link_ops __read_mostly = { | |
1118 | .kind = "gre", | |
1119 | .maxtype = IFLA_GRE_MAX, | |
1120 | .policy = ipgre_policy, | |
1121 | .priv_size = sizeof(struct ip_tunnel), | |
1122 | .setup = ipgre_tunnel_setup, | |
1123 | .validate = ipgre_tunnel_validate, | |
1124 | .newlink = ipgre_newlink, | |
1125 | .changelink = ipgre_changelink, | |
c5441932 | 1126 | .dellink = ip_tunnel_dellink, |
c19e654d HX |
1127 | .get_size = ipgre_get_size, |
1128 | .fill_info = ipgre_fill_info, | |
1728d4fa | 1129 | .get_link_net = ip_tunnel_get_link_net, |
c19e654d HX |
1130 | }; |
1131 | ||
e1a80002 HX |
1132 | static struct rtnl_link_ops ipgre_tap_ops __read_mostly = { |
1133 | .kind = "gretap", | |
1134 | .maxtype = IFLA_GRE_MAX, | |
1135 | .policy = ipgre_policy, | |
1136 | .priv_size = sizeof(struct ip_tunnel), | |
1137 | .setup = ipgre_tap_setup, | |
1138 | .validate = ipgre_tap_validate, | |
1139 | .newlink = ipgre_newlink, | |
1140 | .changelink = ipgre_changelink, | |
c5441932 | 1141 | .dellink = ip_tunnel_dellink, |
e1a80002 HX |
1142 | .get_size = ipgre_get_size, |
1143 | .fill_info = ipgre_fill_info, | |
1728d4fa | 1144 | .get_link_net = ip_tunnel_get_link_net, |
e1a80002 HX |
1145 | }; |
1146 | ||
b2acd1dc PS |
1147 | struct net_device *gretap_fb_dev_create(struct net *net, const char *name, |
1148 | u8 name_assign_type) | |
1149 | { | |
1150 | struct nlattr *tb[IFLA_MAX + 1]; | |
1151 | struct net_device *dev; | |
106da663 | 1152 | LIST_HEAD(list_kill); |
b2acd1dc PS |
1153 | struct ip_tunnel *t; |
1154 | int err; | |
1155 | ||
1156 | memset(&tb, 0, sizeof(tb)); | |
1157 | ||
1158 | dev = rtnl_create_link(net, name, name_assign_type, | |
1159 | &ipgre_tap_ops, tb); | |
1160 | if (IS_ERR(dev)) | |
1161 | return dev; | |
1162 | ||
1163 | /* Configure flow based GRE device. */ | |
1164 | t = netdev_priv(dev); | |
1165 | t->collect_md = true; | |
1166 | ||
1167 | err = ipgre_newlink(net, dev, tb, NULL); | |
106da663 ND |
1168 | if (err < 0) { |
1169 | free_netdev(dev); | |
1170 | return ERR_PTR(err); | |
1171 | } | |
7e059158 DW |
1172 | |
1173 | /* openvswitch users expect packet sizes to be unrestricted, | |
1174 | * so set the largest MTU we can. | |
1175 | */ | |
1176 | err = __ip_tunnel_change_mtu(dev, IP_MAX_MTU, false); | |
1177 | if (err) | |
1178 | goto out; | |
1179 | ||
da6f1da8 ND |
1180 | err = rtnl_configure_link(dev, NULL); |
1181 | if (err < 0) | |
1182 | goto out; | |
1183 | ||
b2acd1dc PS |
1184 | return dev; |
1185 | out: | |
106da663 ND |
1186 | ip_tunnel_dellink(dev, &list_kill); |
1187 | unregister_netdevice_many(&list_kill); | |
b2acd1dc PS |
1188 | return ERR_PTR(err); |
1189 | } | |
1190 | EXPORT_SYMBOL_GPL(gretap_fb_dev_create); | |
1191 | ||
c5441932 PS |
1192 | static int __net_init ipgre_tap_init_net(struct net *net) |
1193 | { | |
2e15ea39 | 1194 | return ip_tunnel_init_net(net, gre_tap_net_id, &ipgre_tap_ops, "gretap0"); |
c5441932 PS |
1195 | } |
1196 | ||
1197 | static void __net_exit ipgre_tap_exit_net(struct net *net) | |
1198 | { | |
1199 | struct ip_tunnel_net *itn = net_generic(net, gre_tap_net_id); | |
6c742e71 | 1200 | ip_tunnel_delete_net(itn, &ipgre_tap_ops); |
c5441932 PS |
1201 | } |
1202 | ||
1203 | static struct pernet_operations ipgre_tap_net_ops = { | |
1204 | .init = ipgre_tap_init_net, | |
1205 | .exit = ipgre_tap_exit_net, | |
1206 | .id = &gre_tap_net_id, | |
1207 | .size = sizeof(struct ip_tunnel_net), | |
1208 | }; | |
1da177e4 LT |
1209 | |
1210 | static int __init ipgre_init(void) | |
1211 | { | |
1212 | int err; | |
1213 | ||
058bd4d2 | 1214 | pr_info("GRE over IPv4 tunneling driver\n"); |
1da177e4 | 1215 | |
cfb8fbf2 | 1216 | err = register_pernet_device(&ipgre_net_ops); |
59a4c759 | 1217 | if (err < 0) |
c2892f02 AD |
1218 | return err; |
1219 | ||
c5441932 PS |
1220 | err = register_pernet_device(&ipgre_tap_net_ops); |
1221 | if (err < 0) | |
1222 | goto pnet_tap_faied; | |
1223 | ||
9f57c67c | 1224 | err = gre_add_protocol(&ipgre_protocol, GREPROTO_CISCO); |
c2892f02 | 1225 | if (err < 0) { |
058bd4d2 | 1226 | pr_info("%s: can't add protocol\n", __func__); |
c2892f02 AD |
1227 | goto add_proto_failed; |
1228 | } | |
7daa0004 | 1229 | |
c19e654d HX |
1230 | err = rtnl_link_register(&ipgre_link_ops); |
1231 | if (err < 0) | |
1232 | goto rtnl_link_failed; | |
1233 | ||
e1a80002 HX |
1234 | err = rtnl_link_register(&ipgre_tap_ops); |
1235 | if (err < 0) | |
1236 | goto tap_ops_failed; | |
1237 | ||
c5441932 | 1238 | return 0; |
c19e654d | 1239 | |
e1a80002 HX |
1240 | tap_ops_failed: |
1241 | rtnl_link_unregister(&ipgre_link_ops); | |
c19e654d | 1242 | rtnl_link_failed: |
9f57c67c | 1243 | gre_del_protocol(&ipgre_protocol, GREPROTO_CISCO); |
c2892f02 | 1244 | add_proto_failed: |
c5441932 PS |
1245 | unregister_pernet_device(&ipgre_tap_net_ops); |
1246 | pnet_tap_faied: | |
c2892f02 | 1247 | unregister_pernet_device(&ipgre_net_ops); |
c5441932 | 1248 | return err; |
1da177e4 LT |
1249 | } |
1250 | ||
db44575f | 1251 | static void __exit ipgre_fini(void) |
1da177e4 | 1252 | { |
e1a80002 | 1253 | rtnl_link_unregister(&ipgre_tap_ops); |
c19e654d | 1254 | rtnl_link_unregister(&ipgre_link_ops); |
9f57c67c | 1255 | gre_del_protocol(&ipgre_protocol, GREPROTO_CISCO); |
c5441932 | 1256 | unregister_pernet_device(&ipgre_tap_net_ops); |
c2892f02 | 1257 | unregister_pernet_device(&ipgre_net_ops); |
1da177e4 LT |
1258 | } |
1259 | ||
1260 | module_init(ipgre_init); | |
1261 | module_exit(ipgre_fini); | |
1262 | MODULE_LICENSE("GPL"); | |
4d74f8ba PM |
1263 | MODULE_ALIAS_RTNL_LINK("gre"); |
1264 | MODULE_ALIAS_RTNL_LINK("gretap"); | |
8909c9ad | 1265 | MODULE_ALIAS_NETDEV("gre0"); |
c5441932 | 1266 | MODULE_ALIAS_NETDEV("gretap0"); |