xref: /openbmc/linux/net/netfilter/nft_tunnel.c (revision 5334d983)
1 /* SPDX-License-Identifier: GPL-2.0 */
2 #include <linux/kernel.h>
3 #include <linux/init.h>
4 #include <linux/module.h>
5 #include <linux/seqlock.h>
6 #include <linux/netlink.h>
7 #include <linux/netfilter.h>
8 #include <linux/netfilter/nf_tables.h>
9 #include <net/netfilter/nf_tables.h>
10 #include <net/dst_metadata.h>
11 #include <net/ip_tunnels.h>
12 #include <net/vxlan.h>
13 #include <net/erspan.h>
14 #include <net/geneve.h>
15 
16 struct nft_tunnel {
17 	enum nft_tunnel_keys	key:8;
18 	u8			dreg;
19 	enum nft_tunnel_mode	mode:8;
20 };
21 
22 static void nft_tunnel_get_eval(const struct nft_expr *expr,
23 				struct nft_regs *regs,
24 				const struct nft_pktinfo *pkt)
25 {
26 	const struct nft_tunnel *priv = nft_expr_priv(expr);
27 	u32 *dest = &regs->data[priv->dreg];
28 	struct ip_tunnel_info *tun_info;
29 
30 	tun_info = skb_tunnel_info(pkt->skb);
31 
32 	switch (priv->key) {
33 	case NFT_TUNNEL_PATH:
34 		if (!tun_info) {
35 			nft_reg_store8(dest, false);
36 			return;
37 		}
38 		if (priv->mode == NFT_TUNNEL_MODE_NONE ||
39 		    (priv->mode == NFT_TUNNEL_MODE_RX &&
40 		     !(tun_info->mode & IP_TUNNEL_INFO_TX)) ||
41 		    (priv->mode == NFT_TUNNEL_MODE_TX &&
42 		     (tun_info->mode & IP_TUNNEL_INFO_TX)))
43 			nft_reg_store8(dest, true);
44 		else
45 			nft_reg_store8(dest, false);
46 		break;
47 	case NFT_TUNNEL_ID:
48 		if (!tun_info) {
49 			regs->verdict.code = NFT_BREAK;
50 			return;
51 		}
52 		if (priv->mode == NFT_TUNNEL_MODE_NONE ||
53 		    (priv->mode == NFT_TUNNEL_MODE_RX &&
54 		     !(tun_info->mode & IP_TUNNEL_INFO_TX)) ||
55 		    (priv->mode == NFT_TUNNEL_MODE_TX &&
56 		     (tun_info->mode & IP_TUNNEL_INFO_TX)))
57 			*dest = ntohl(tunnel_id_to_key32(tun_info->key.tun_id));
58 		else
59 			regs->verdict.code = NFT_BREAK;
60 		break;
61 	default:
62 		WARN_ON(1);
63 		regs->verdict.code = NFT_BREAK;
64 	}
65 }
66 
67 static const struct nla_policy nft_tunnel_policy[NFTA_TUNNEL_MAX + 1] = {
68 	[NFTA_TUNNEL_KEY]	= { .type = NLA_U32 },
69 	[NFTA_TUNNEL_DREG]	= { .type = NLA_U32 },
70 	[NFTA_TUNNEL_MODE]	= { .type = NLA_U32 },
71 };
72 
73 static int nft_tunnel_get_init(const struct nft_ctx *ctx,
74 			       const struct nft_expr *expr,
75 			       const struct nlattr * const tb[])
76 {
77 	struct nft_tunnel *priv = nft_expr_priv(expr);
78 	u32 len;
79 
80 	if (!tb[NFTA_TUNNEL_KEY] ||
81 	    !tb[NFTA_TUNNEL_DREG])
82 		return -EINVAL;
83 
84 	priv->key = ntohl(nla_get_be32(tb[NFTA_TUNNEL_KEY]));
85 	switch (priv->key) {
86 	case NFT_TUNNEL_PATH:
87 		len = sizeof(u8);
88 		break;
89 	case NFT_TUNNEL_ID:
90 		len = sizeof(u32);
91 		break;
92 	default:
93 		return -EOPNOTSUPP;
94 	}
95 
96 	if (tb[NFTA_TUNNEL_MODE]) {
97 		priv->mode = ntohl(nla_get_be32(tb[NFTA_TUNNEL_MODE]));
98 		if (priv->mode > NFT_TUNNEL_MODE_MAX)
99 			return -EOPNOTSUPP;
100 	} else {
101 		priv->mode = NFT_TUNNEL_MODE_NONE;
102 	}
103 
104 	return nft_parse_register_store(ctx, tb[NFTA_TUNNEL_DREG], &priv->dreg,
105 					NULL, NFT_DATA_VALUE, len);
106 }
107 
108 static int nft_tunnel_get_dump(struct sk_buff *skb,
109 			       const struct nft_expr *expr)
110 {
111 	const struct nft_tunnel *priv = nft_expr_priv(expr);
112 
113 	if (nla_put_be32(skb, NFTA_TUNNEL_KEY, htonl(priv->key)))
114 		goto nla_put_failure;
115 	if (nft_dump_register(skb, NFTA_TUNNEL_DREG, priv->dreg))
116 		goto nla_put_failure;
117 	if (nla_put_be32(skb, NFTA_TUNNEL_MODE, htonl(priv->mode)))
118 		goto nla_put_failure;
119 	return 0;
120 
121 nla_put_failure:
122 	return -1;
123 }
124 
125 static struct nft_expr_type nft_tunnel_type;
126 static const struct nft_expr_ops nft_tunnel_get_ops = {
127 	.type		= &nft_tunnel_type,
128 	.size		= NFT_EXPR_SIZE(sizeof(struct nft_tunnel)),
129 	.eval		= nft_tunnel_get_eval,
130 	.init		= nft_tunnel_get_init,
131 	.dump		= nft_tunnel_get_dump,
132 };
133 
134 static struct nft_expr_type nft_tunnel_type __read_mostly = {
135 	.name		= "tunnel",
136 	.ops		= &nft_tunnel_get_ops,
137 	.policy		= nft_tunnel_policy,
138 	.maxattr	= NFTA_TUNNEL_MAX,
139 	.owner		= THIS_MODULE,
140 };
141 
142 struct nft_tunnel_opts {
143 	union {
144 		struct vxlan_metadata	vxlan;
145 		struct erspan_metadata	erspan;
146 		u8	data[IP_TUNNEL_OPTS_MAX];
147 	} u;
148 	u32	len;
149 	__be16	flags;
150 };
151 
152 struct nft_tunnel_obj {
153 	struct metadata_dst	*md;
154 	struct nft_tunnel_opts	opts;
155 };
156 
157 static const struct nla_policy nft_tunnel_ip_policy[NFTA_TUNNEL_KEY_IP_MAX + 1] = {
158 	[NFTA_TUNNEL_KEY_IP_SRC]	= { .type = NLA_U32 },
159 	[NFTA_TUNNEL_KEY_IP_DST]	= { .type = NLA_U32 },
160 };
161 
162 static int nft_tunnel_obj_ip_init(const struct nft_ctx *ctx,
163 				  const struct nlattr *attr,
164 				  struct ip_tunnel_info *info)
165 {
166 	struct nlattr *tb[NFTA_TUNNEL_KEY_IP_MAX + 1];
167 	int err;
168 
169 	err = nla_parse_nested_deprecated(tb, NFTA_TUNNEL_KEY_IP_MAX, attr,
170 					  nft_tunnel_ip_policy, NULL);
171 	if (err < 0)
172 		return err;
173 
174 	if (!tb[NFTA_TUNNEL_KEY_IP_DST])
175 		return -EINVAL;
176 
177 	if (tb[NFTA_TUNNEL_KEY_IP_SRC])
178 		info->key.u.ipv4.src = nla_get_be32(tb[NFTA_TUNNEL_KEY_IP_SRC]);
179 	if (tb[NFTA_TUNNEL_KEY_IP_DST])
180 		info->key.u.ipv4.dst = nla_get_be32(tb[NFTA_TUNNEL_KEY_IP_DST]);
181 
182 	return 0;
183 }
184 
185 static const struct nla_policy nft_tunnel_ip6_policy[NFTA_TUNNEL_KEY_IP6_MAX + 1] = {
186 	[NFTA_TUNNEL_KEY_IP6_SRC]	= { .len = sizeof(struct in6_addr), },
187 	[NFTA_TUNNEL_KEY_IP6_DST]	= { .len = sizeof(struct in6_addr), },
188 	[NFTA_TUNNEL_KEY_IP6_FLOWLABEL]	= { .type = NLA_U32, }
189 };
190 
191 static int nft_tunnel_obj_ip6_init(const struct nft_ctx *ctx,
192 				   const struct nlattr *attr,
193 				   struct ip_tunnel_info *info)
194 {
195 	struct nlattr *tb[NFTA_TUNNEL_KEY_IP6_MAX + 1];
196 	int err;
197 
198 	err = nla_parse_nested_deprecated(tb, NFTA_TUNNEL_KEY_IP6_MAX, attr,
199 					  nft_tunnel_ip6_policy, NULL);
200 	if (err < 0)
201 		return err;
202 
203 	if (!tb[NFTA_TUNNEL_KEY_IP6_DST])
204 		return -EINVAL;
205 
206 	if (tb[NFTA_TUNNEL_KEY_IP6_SRC]) {
207 		memcpy(&info->key.u.ipv6.src,
208 		       nla_data(tb[NFTA_TUNNEL_KEY_IP6_SRC]),
209 		       sizeof(struct in6_addr));
210 	}
211 	if (tb[NFTA_TUNNEL_KEY_IP6_DST]) {
212 		memcpy(&info->key.u.ipv6.dst,
213 		       nla_data(tb[NFTA_TUNNEL_KEY_IP6_DST]),
214 		       sizeof(struct in6_addr));
215 	}
216 	if (tb[NFTA_TUNNEL_KEY_IP6_FLOWLABEL])
217 		info->key.label = nla_get_be32(tb[NFTA_TUNNEL_KEY_IP6_FLOWLABEL]);
218 
219 	info->mode |= IP_TUNNEL_INFO_IPV6;
220 
221 	return 0;
222 }
223 
224 static const struct nla_policy nft_tunnel_opts_vxlan_policy[NFTA_TUNNEL_KEY_VXLAN_MAX + 1] = {
225 	[NFTA_TUNNEL_KEY_VXLAN_GBP]	= { .type = NLA_U32 },
226 };
227 
228 static int nft_tunnel_obj_vxlan_init(const struct nlattr *attr,
229 				     struct nft_tunnel_opts *opts)
230 {
231 	struct nlattr *tb[NFTA_TUNNEL_KEY_VXLAN_MAX + 1];
232 	int err;
233 
234 	err = nla_parse_nested_deprecated(tb, NFTA_TUNNEL_KEY_VXLAN_MAX, attr,
235 					  nft_tunnel_opts_vxlan_policy, NULL);
236 	if (err < 0)
237 		return err;
238 
239 	if (!tb[NFTA_TUNNEL_KEY_VXLAN_GBP])
240 		return -EINVAL;
241 
242 	opts->u.vxlan.gbp = ntohl(nla_get_be32(tb[NFTA_TUNNEL_KEY_VXLAN_GBP]));
243 
244 	opts->len	= sizeof(struct vxlan_metadata);
245 	opts->flags	= TUNNEL_VXLAN_OPT;
246 
247 	return 0;
248 }
249 
250 static const struct nla_policy nft_tunnel_opts_erspan_policy[NFTA_TUNNEL_KEY_ERSPAN_MAX + 1] = {
251 	[NFTA_TUNNEL_KEY_ERSPAN_VERSION]	= { .type = NLA_U32 },
252 	[NFTA_TUNNEL_KEY_ERSPAN_V1_INDEX]	= { .type = NLA_U32 },
253 	[NFTA_TUNNEL_KEY_ERSPAN_V2_DIR]		= { .type = NLA_U8 },
254 	[NFTA_TUNNEL_KEY_ERSPAN_V2_HWID]	= { .type = NLA_U8 },
255 };
256 
257 static int nft_tunnel_obj_erspan_init(const struct nlattr *attr,
258 				      struct nft_tunnel_opts *opts)
259 {
260 	struct nlattr *tb[NFTA_TUNNEL_KEY_ERSPAN_MAX + 1];
261 	uint8_t hwid, dir;
262 	int err, version;
263 
264 	err = nla_parse_nested_deprecated(tb, NFTA_TUNNEL_KEY_ERSPAN_MAX,
265 					  attr, nft_tunnel_opts_erspan_policy,
266 					  NULL);
267 	if (err < 0)
268 		return err;
269 
270 	if (!tb[NFTA_TUNNEL_KEY_ERSPAN_VERSION])
271 		 return -EINVAL;
272 
273 	version = ntohl(nla_get_be32(tb[NFTA_TUNNEL_KEY_ERSPAN_VERSION]));
274 	switch (version) {
275 	case ERSPAN_VERSION:
276 		if (!tb[NFTA_TUNNEL_KEY_ERSPAN_V1_INDEX])
277 			return -EINVAL;
278 
279 		opts->u.erspan.u.index =
280 			nla_get_be32(tb[NFTA_TUNNEL_KEY_ERSPAN_V1_INDEX]);
281 		break;
282 	case ERSPAN_VERSION2:
283 		if (!tb[NFTA_TUNNEL_KEY_ERSPAN_V2_DIR] ||
284 		    !tb[NFTA_TUNNEL_KEY_ERSPAN_V2_HWID])
285 			return -EINVAL;
286 
287 		hwid = nla_get_u8(tb[NFTA_TUNNEL_KEY_ERSPAN_V2_HWID]);
288 		dir = nla_get_u8(tb[NFTA_TUNNEL_KEY_ERSPAN_V2_DIR]);
289 
290 		set_hwid(&opts->u.erspan.u.md2, hwid);
291 		opts->u.erspan.u.md2.dir = dir;
292 		break;
293 	default:
294 		return -EOPNOTSUPP;
295 	}
296 	opts->u.erspan.version = version;
297 
298 	opts->len	= sizeof(struct erspan_metadata);
299 	opts->flags	= TUNNEL_ERSPAN_OPT;
300 
301 	return 0;
302 }
303 
304 static const struct nla_policy nft_tunnel_opts_geneve_policy[NFTA_TUNNEL_KEY_GENEVE_MAX + 1] = {
305 	[NFTA_TUNNEL_KEY_GENEVE_CLASS]	= { .type = NLA_U16 },
306 	[NFTA_TUNNEL_KEY_GENEVE_TYPE]	= { .type = NLA_U8 },
307 	[NFTA_TUNNEL_KEY_GENEVE_DATA]	= { .type = NLA_BINARY, .len = 128 },
308 };
309 
310 static int nft_tunnel_obj_geneve_init(const struct nlattr *attr,
311 				      struct nft_tunnel_opts *opts)
312 {
313 	struct geneve_opt *opt = (struct geneve_opt *)opts->u.data + opts->len;
314 	struct nlattr *tb[NFTA_TUNNEL_KEY_GENEVE_MAX + 1];
315 	int err, data_len;
316 
317 	err = nla_parse_nested(tb, NFTA_TUNNEL_KEY_GENEVE_MAX, attr,
318 			       nft_tunnel_opts_geneve_policy, NULL);
319 	if (err < 0)
320 		return err;
321 
322 	if (!tb[NFTA_TUNNEL_KEY_GENEVE_CLASS] ||
323 	    !tb[NFTA_TUNNEL_KEY_GENEVE_TYPE] ||
324 	    !tb[NFTA_TUNNEL_KEY_GENEVE_DATA])
325 		return -EINVAL;
326 
327 	attr = tb[NFTA_TUNNEL_KEY_GENEVE_DATA];
328 	data_len = nla_len(attr);
329 	if (data_len % 4)
330 		return -EINVAL;
331 
332 	opts->len += sizeof(*opt) + data_len;
333 	if (opts->len > IP_TUNNEL_OPTS_MAX)
334 		return -EINVAL;
335 
336 	memcpy(opt->opt_data, nla_data(attr), data_len);
337 	opt->length = data_len / 4;
338 	opt->opt_class = nla_get_be16(tb[NFTA_TUNNEL_KEY_GENEVE_CLASS]);
339 	opt->type = nla_get_u8(tb[NFTA_TUNNEL_KEY_GENEVE_TYPE]);
340 	opts->flags = TUNNEL_GENEVE_OPT;
341 
342 	return 0;
343 }
344 
345 static const struct nla_policy nft_tunnel_opts_policy[NFTA_TUNNEL_KEY_OPTS_MAX + 1] = {
346 	[NFTA_TUNNEL_KEY_OPTS_UNSPEC]	= {
347 		.strict_start_type = NFTA_TUNNEL_KEY_OPTS_GENEVE },
348 	[NFTA_TUNNEL_KEY_OPTS_VXLAN]	= { .type = NLA_NESTED, },
349 	[NFTA_TUNNEL_KEY_OPTS_ERSPAN]	= { .type = NLA_NESTED, },
350 	[NFTA_TUNNEL_KEY_OPTS_GENEVE]	= { .type = NLA_NESTED, },
351 };
352 
353 static int nft_tunnel_obj_opts_init(const struct nft_ctx *ctx,
354 				    const struct nlattr *attr,
355 				    struct ip_tunnel_info *info,
356 				    struct nft_tunnel_opts *opts)
357 {
358 	int err, rem, type = 0;
359 	struct nlattr *nla;
360 
361 	err = nla_validate_nested_deprecated(attr, NFTA_TUNNEL_KEY_OPTS_MAX,
362 					     nft_tunnel_opts_policy, NULL);
363 	if (err < 0)
364 		return err;
365 
366 	nla_for_each_attr(nla, nla_data(attr), nla_len(attr), rem) {
367 		switch (nla_type(nla)) {
368 		case NFTA_TUNNEL_KEY_OPTS_VXLAN:
369 			if (type)
370 				return -EINVAL;
371 			err = nft_tunnel_obj_vxlan_init(nla, opts);
372 			if (err)
373 				return err;
374 			type = TUNNEL_VXLAN_OPT;
375 			break;
376 		case NFTA_TUNNEL_KEY_OPTS_ERSPAN:
377 			if (type)
378 				return -EINVAL;
379 			err = nft_tunnel_obj_erspan_init(nla, opts);
380 			if (err)
381 				return err;
382 			type = TUNNEL_ERSPAN_OPT;
383 			break;
384 		case NFTA_TUNNEL_KEY_OPTS_GENEVE:
385 			if (type && type != TUNNEL_GENEVE_OPT)
386 				return -EINVAL;
387 			err = nft_tunnel_obj_geneve_init(nla, opts);
388 			if (err)
389 				return err;
390 			type = TUNNEL_GENEVE_OPT;
391 			break;
392 		default:
393 			return -EOPNOTSUPP;
394 		}
395 	}
396 
397 	return err;
398 }
399 
400 static const struct nla_policy nft_tunnel_key_policy[NFTA_TUNNEL_KEY_MAX + 1] = {
401 	[NFTA_TUNNEL_KEY_IP]	= { .type = NLA_NESTED, },
402 	[NFTA_TUNNEL_KEY_IP6]	= { .type = NLA_NESTED, },
403 	[NFTA_TUNNEL_KEY_ID]	= { .type = NLA_U32, },
404 	[NFTA_TUNNEL_KEY_FLAGS]	= { .type = NLA_U32, },
405 	[NFTA_TUNNEL_KEY_TOS]	= { .type = NLA_U8, },
406 	[NFTA_TUNNEL_KEY_TTL]	= { .type = NLA_U8, },
407 	[NFTA_TUNNEL_KEY_SPORT]	= { .type = NLA_U16, },
408 	[NFTA_TUNNEL_KEY_DPORT]	= { .type = NLA_U16, },
409 	[NFTA_TUNNEL_KEY_OPTS]	= { .type = NLA_NESTED, },
410 };
411 
412 static int nft_tunnel_obj_init(const struct nft_ctx *ctx,
413 			       const struct nlattr * const tb[],
414 			       struct nft_object *obj)
415 {
416 	struct nft_tunnel_obj *priv = nft_obj_data(obj);
417 	struct ip_tunnel_info info;
418 	struct metadata_dst *md;
419 	int err;
420 
421 	if (!tb[NFTA_TUNNEL_KEY_ID])
422 		return -EINVAL;
423 
424 	memset(&info, 0, sizeof(info));
425 	info.mode		= IP_TUNNEL_INFO_TX;
426 	info.key.tun_id		= key32_to_tunnel_id(nla_get_be32(tb[NFTA_TUNNEL_KEY_ID]));
427 	info.key.tun_flags	= TUNNEL_KEY | TUNNEL_CSUM | TUNNEL_NOCACHE;
428 
429 	if (tb[NFTA_TUNNEL_KEY_IP]) {
430 		err = nft_tunnel_obj_ip_init(ctx, tb[NFTA_TUNNEL_KEY_IP], &info);
431 		if (err < 0)
432 			return err;
433 	} else if (tb[NFTA_TUNNEL_KEY_IP6]) {
434 		err = nft_tunnel_obj_ip6_init(ctx, tb[NFTA_TUNNEL_KEY_IP6], &info);
435 		if (err < 0)
436 			return err;
437 	} else {
438 		return -EINVAL;
439 	}
440 
441 	if (tb[NFTA_TUNNEL_KEY_SPORT]) {
442 		info.key.tp_src = nla_get_be16(tb[NFTA_TUNNEL_KEY_SPORT]);
443 	}
444 	if (tb[NFTA_TUNNEL_KEY_DPORT]) {
445 		info.key.tp_dst = nla_get_be16(tb[NFTA_TUNNEL_KEY_DPORT]);
446 	}
447 
448 	if (tb[NFTA_TUNNEL_KEY_FLAGS]) {
449 		u32 tun_flags;
450 
451 		tun_flags = ntohl(nla_get_be32(tb[NFTA_TUNNEL_KEY_FLAGS]));
452 		if (tun_flags & ~NFT_TUNNEL_F_MASK)
453 			return -EOPNOTSUPP;
454 
455 		if (tun_flags & NFT_TUNNEL_F_ZERO_CSUM_TX)
456 			info.key.tun_flags &= ~TUNNEL_CSUM;
457 		if (tun_flags & NFT_TUNNEL_F_DONT_FRAGMENT)
458 			info.key.tun_flags |= TUNNEL_DONT_FRAGMENT;
459 		if (tun_flags & NFT_TUNNEL_F_SEQ_NUMBER)
460 			info.key.tun_flags |= TUNNEL_SEQ;
461 	}
462 	if (tb[NFTA_TUNNEL_KEY_TOS])
463 		info.key.tos = nla_get_u8(tb[NFTA_TUNNEL_KEY_TOS]);
464 	if (tb[NFTA_TUNNEL_KEY_TTL])
465 		info.key.ttl = nla_get_u8(tb[NFTA_TUNNEL_KEY_TTL]);
466 	else
467 		info.key.ttl = U8_MAX;
468 
469 	if (tb[NFTA_TUNNEL_KEY_OPTS]) {
470 		err = nft_tunnel_obj_opts_init(ctx, tb[NFTA_TUNNEL_KEY_OPTS],
471 					       &info, &priv->opts);
472 		if (err < 0)
473 			return err;
474 	}
475 
476 	md = metadata_dst_alloc(priv->opts.len, METADATA_IP_TUNNEL, GFP_KERNEL);
477 	if (!md)
478 		return -ENOMEM;
479 
480 	memcpy(&md->u.tun_info, &info, sizeof(info));
481 #ifdef CONFIG_DST_CACHE
482 	err = dst_cache_init(&md->u.tun_info.dst_cache, GFP_KERNEL);
483 	if (err < 0) {
484 		metadata_dst_free(md);
485 		return err;
486 	}
487 #endif
488 	ip_tunnel_info_opts_set(&md->u.tun_info, &priv->opts.u, priv->opts.len,
489 				priv->opts.flags);
490 	priv->md = md;
491 
492 	return 0;
493 }
494 
495 static inline void nft_tunnel_obj_eval(struct nft_object *obj,
496 				       struct nft_regs *regs,
497 				       const struct nft_pktinfo *pkt)
498 {
499 	struct nft_tunnel_obj *priv = nft_obj_data(obj);
500 	struct sk_buff *skb = pkt->skb;
501 
502 	skb_dst_drop(skb);
503 	dst_hold((struct dst_entry *) priv->md);
504 	skb_dst_set(skb, (struct dst_entry *) priv->md);
505 }
506 
507 static int nft_tunnel_ip_dump(struct sk_buff *skb, struct ip_tunnel_info *info)
508 {
509 	struct nlattr *nest;
510 
511 	if (info->mode & IP_TUNNEL_INFO_IPV6) {
512 		nest = nla_nest_start_noflag(skb, NFTA_TUNNEL_KEY_IP6);
513 		if (!nest)
514 			return -1;
515 
516 		if (nla_put_in6_addr(skb, NFTA_TUNNEL_KEY_IP6_SRC,
517 				     &info->key.u.ipv6.src) < 0 ||
518 		    nla_put_in6_addr(skb, NFTA_TUNNEL_KEY_IP6_DST,
519 				     &info->key.u.ipv6.dst) < 0 ||
520 		    nla_put_be32(skb, NFTA_TUNNEL_KEY_IP6_FLOWLABEL,
521 				 info->key.label)) {
522 			nla_nest_cancel(skb, nest);
523 			return -1;
524 		}
525 
526 		nla_nest_end(skb, nest);
527 	} else {
528 		nest = nla_nest_start_noflag(skb, NFTA_TUNNEL_KEY_IP);
529 		if (!nest)
530 			return -1;
531 
532 		if (nla_put_in_addr(skb, NFTA_TUNNEL_KEY_IP_SRC,
533 				    info->key.u.ipv4.src) < 0 ||
534 		    nla_put_in_addr(skb, NFTA_TUNNEL_KEY_IP_DST,
535 				    info->key.u.ipv4.dst) < 0) {
536 			nla_nest_cancel(skb, nest);
537 			return -1;
538 		}
539 
540 		nla_nest_end(skb, nest);
541 	}
542 
543 	return 0;
544 }
545 
546 static int nft_tunnel_opts_dump(struct sk_buff *skb,
547 				struct nft_tunnel_obj *priv)
548 {
549 	struct nft_tunnel_opts *opts = &priv->opts;
550 	struct nlattr *nest, *inner;
551 
552 	nest = nla_nest_start_noflag(skb, NFTA_TUNNEL_KEY_OPTS);
553 	if (!nest)
554 		return -1;
555 
556 	if (opts->flags & TUNNEL_VXLAN_OPT) {
557 		inner = nla_nest_start_noflag(skb, NFTA_TUNNEL_KEY_OPTS_VXLAN);
558 		if (!inner)
559 			goto failure;
560 		if (nla_put_be32(skb, NFTA_TUNNEL_KEY_VXLAN_GBP,
561 				 htonl(opts->u.vxlan.gbp)))
562 			goto inner_failure;
563 		nla_nest_end(skb, inner);
564 	} else if (opts->flags & TUNNEL_ERSPAN_OPT) {
565 		inner = nla_nest_start_noflag(skb, NFTA_TUNNEL_KEY_OPTS_ERSPAN);
566 		if (!inner)
567 			goto failure;
568 		if (nla_put_be32(skb, NFTA_TUNNEL_KEY_ERSPAN_VERSION,
569 				 htonl(opts->u.erspan.version)))
570 			goto inner_failure;
571 		switch (opts->u.erspan.version) {
572 		case ERSPAN_VERSION:
573 			if (nla_put_be32(skb, NFTA_TUNNEL_KEY_ERSPAN_V1_INDEX,
574 					 opts->u.erspan.u.index))
575 				goto inner_failure;
576 			break;
577 		case ERSPAN_VERSION2:
578 			if (nla_put_u8(skb, NFTA_TUNNEL_KEY_ERSPAN_V2_HWID,
579 				       get_hwid(&opts->u.erspan.u.md2)) ||
580 			    nla_put_u8(skb, NFTA_TUNNEL_KEY_ERSPAN_V2_DIR,
581 				       opts->u.erspan.u.md2.dir))
582 				goto inner_failure;
583 			break;
584 		}
585 		nla_nest_end(skb, inner);
586 	} else if (opts->flags & TUNNEL_GENEVE_OPT) {
587 		struct geneve_opt *opt;
588 		int offset = 0;
589 
590 		inner = nla_nest_start_noflag(skb, NFTA_TUNNEL_KEY_OPTS_GENEVE);
591 		if (!inner)
592 			goto failure;
593 		while (opts->len > offset) {
594 			opt = (struct geneve_opt *)opts->u.data + offset;
595 			if (nla_put_be16(skb, NFTA_TUNNEL_KEY_GENEVE_CLASS,
596 					 opt->opt_class) ||
597 			    nla_put_u8(skb, NFTA_TUNNEL_KEY_GENEVE_TYPE,
598 				       opt->type) ||
599 			    nla_put(skb, NFTA_TUNNEL_KEY_GENEVE_DATA,
600 				    opt->length * 4, opt->opt_data))
601 				goto inner_failure;
602 			offset += sizeof(*opt) + opt->length * 4;
603 		}
604 		nla_nest_end(skb, inner);
605 	}
606 	nla_nest_end(skb, nest);
607 	return 0;
608 
609 inner_failure:
610 	nla_nest_cancel(skb, inner);
611 failure:
612 	nla_nest_cancel(skb, nest);
613 	return -1;
614 }
615 
616 static int nft_tunnel_ports_dump(struct sk_buff *skb,
617 				 struct ip_tunnel_info *info)
618 {
619 	if (nla_put_be16(skb, NFTA_TUNNEL_KEY_SPORT, info->key.tp_src) < 0 ||
620 	    nla_put_be16(skb, NFTA_TUNNEL_KEY_DPORT, info->key.tp_dst) < 0)
621 		return -1;
622 
623 	return 0;
624 }
625 
626 static int nft_tunnel_flags_dump(struct sk_buff *skb,
627 				 struct ip_tunnel_info *info)
628 {
629 	u32 flags = 0;
630 
631 	if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT)
632 		flags |= NFT_TUNNEL_F_DONT_FRAGMENT;
633 	if (!(info->key.tun_flags & TUNNEL_CSUM))
634 		flags |= NFT_TUNNEL_F_ZERO_CSUM_TX;
635 	if (info->key.tun_flags & TUNNEL_SEQ)
636 		flags |= NFT_TUNNEL_F_SEQ_NUMBER;
637 
638 	if (nla_put_be32(skb, NFTA_TUNNEL_KEY_FLAGS, htonl(flags)) < 0)
639 		return -1;
640 
641 	return 0;
642 }
643 
644 static int nft_tunnel_obj_dump(struct sk_buff *skb,
645 			       struct nft_object *obj, bool reset)
646 {
647 	struct nft_tunnel_obj *priv = nft_obj_data(obj);
648 	struct ip_tunnel_info *info = &priv->md->u.tun_info;
649 
650 	if (nla_put_be32(skb, NFTA_TUNNEL_KEY_ID,
651 			 tunnel_id_to_key32(info->key.tun_id)) ||
652 	    nft_tunnel_ip_dump(skb, info) < 0 ||
653 	    nft_tunnel_ports_dump(skb, info) < 0 ||
654 	    nft_tunnel_flags_dump(skb, info) < 0 ||
655 	    nla_put_u8(skb, NFTA_TUNNEL_KEY_TOS, info->key.tos) ||
656 	    nla_put_u8(skb, NFTA_TUNNEL_KEY_TTL, info->key.ttl) ||
657 	    nft_tunnel_opts_dump(skb, priv) < 0)
658 		goto nla_put_failure;
659 
660 	return 0;
661 
662 nla_put_failure:
663 	return -1;
664 }
665 
666 static void nft_tunnel_obj_destroy(const struct nft_ctx *ctx,
667 				   struct nft_object *obj)
668 {
669 	struct nft_tunnel_obj *priv = nft_obj_data(obj);
670 
671 	metadata_dst_free(priv->md);
672 }
673 
674 static struct nft_object_type nft_tunnel_obj_type;
675 static const struct nft_object_ops nft_tunnel_obj_ops = {
676 	.type		= &nft_tunnel_obj_type,
677 	.size		= sizeof(struct nft_tunnel_obj),
678 	.eval		= nft_tunnel_obj_eval,
679 	.init		= nft_tunnel_obj_init,
680 	.destroy	= nft_tunnel_obj_destroy,
681 	.dump		= nft_tunnel_obj_dump,
682 };
683 
684 static struct nft_object_type nft_tunnel_obj_type __read_mostly = {
685 	.type		= NFT_OBJECT_TUNNEL,
686 	.ops		= &nft_tunnel_obj_ops,
687 	.maxattr	= NFTA_TUNNEL_KEY_MAX,
688 	.policy		= nft_tunnel_key_policy,
689 	.owner		= THIS_MODULE,
690 };
691 
692 static int __init nft_tunnel_module_init(void)
693 {
694 	int err;
695 
696 	err = nft_register_expr(&nft_tunnel_type);
697 	if (err < 0)
698 		return err;
699 
700 	err = nft_register_obj(&nft_tunnel_obj_type);
701 	if (err < 0)
702 		nft_unregister_expr(&nft_tunnel_type);
703 
704 	return err;
705 }
706 
707 static void __exit nft_tunnel_module_exit(void)
708 {
709 	nft_unregister_obj(&nft_tunnel_obj_type);
710 	nft_unregister_expr(&nft_tunnel_type);
711 }
712 
713 module_init(nft_tunnel_module_init);
714 module_exit(nft_tunnel_module_exit);
715 
716 MODULE_LICENSE("GPL");
717 MODULE_AUTHOR("Pablo Neira Ayuso <pablo@netfilter.org>");
718 MODULE_ALIAS_NFT_EXPR("tunnel");
719 MODULE_ALIAS_NFT_OBJ(NFT_OBJECT_TUNNEL);
720 MODULE_DESCRIPTION("nftables tunnel expression support");
721