xref: /linux/net/netfilter/nft_tunnel.c (revision b83deaa741558babf4b8d51d34f6637ccfff1b26)
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 	u8			len;
21 };
22 
23 static void nft_tunnel_get_eval(const struct nft_expr *expr,
24 				struct nft_regs *regs,
25 				const struct nft_pktinfo *pkt)
26 {
27 	const struct nft_tunnel *priv = nft_expr_priv(expr);
28 	u32 *dest = &regs->data[priv->dreg];
29 	struct ip_tunnel_info *tun_info;
30 
31 	tun_info = skb_tunnel_info(pkt->skb);
32 
33 	switch (priv->key) {
34 	case NFT_TUNNEL_PATH:
35 		if (!tun_info) {
36 			nft_reg_store8(dest, false);
37 			return;
38 		}
39 		if (priv->mode == NFT_TUNNEL_MODE_NONE ||
40 		    (priv->mode == NFT_TUNNEL_MODE_RX &&
41 		     !(tun_info->mode & IP_TUNNEL_INFO_TX)) ||
42 		    (priv->mode == NFT_TUNNEL_MODE_TX &&
43 		     (tun_info->mode & IP_TUNNEL_INFO_TX)))
44 			nft_reg_store8(dest, true);
45 		else
46 			nft_reg_store8(dest, false);
47 		break;
48 	case NFT_TUNNEL_ID:
49 		if (!tun_info) {
50 			regs->verdict.code = NFT_BREAK;
51 			return;
52 		}
53 		if (priv->mode == NFT_TUNNEL_MODE_NONE ||
54 		    (priv->mode == NFT_TUNNEL_MODE_RX &&
55 		     !(tun_info->mode & IP_TUNNEL_INFO_TX)) ||
56 		    (priv->mode == NFT_TUNNEL_MODE_TX &&
57 		     (tun_info->mode & IP_TUNNEL_INFO_TX)))
58 			*dest = ntohl(tunnel_id_to_key32(tun_info->key.tun_id));
59 		else
60 			regs->verdict.code = NFT_BREAK;
61 		break;
62 	default:
63 		WARN_ON(1);
64 		regs->verdict.code = NFT_BREAK;
65 	}
66 }
67 
68 static const struct nla_policy nft_tunnel_policy[NFTA_TUNNEL_MAX + 1] = {
69 	[NFTA_TUNNEL_KEY]	= { .type = NLA_U32 },
70 	[NFTA_TUNNEL_DREG]	= { .type = NLA_U32 },
71 	[NFTA_TUNNEL_MODE]	= { .type = NLA_U32 },
72 };
73 
74 static int nft_tunnel_get_init(const struct nft_ctx *ctx,
75 			       const struct nft_expr *expr,
76 			       const struct nlattr * const tb[])
77 {
78 	struct nft_tunnel *priv = nft_expr_priv(expr);
79 	u32 len;
80 
81 	if (!tb[NFTA_TUNNEL_KEY] ||
82 	    !tb[NFTA_TUNNEL_DREG])
83 		return -EINVAL;
84 
85 	priv->key = ntohl(nla_get_be32(tb[NFTA_TUNNEL_KEY]));
86 	switch (priv->key) {
87 	case NFT_TUNNEL_PATH:
88 		len = sizeof(u8);
89 		break;
90 	case NFT_TUNNEL_ID:
91 		len = sizeof(u32);
92 		break;
93 	default:
94 		return -EOPNOTSUPP;
95 	}
96 
97 	if (tb[NFTA_TUNNEL_MODE]) {
98 		priv->mode = ntohl(nla_get_be32(tb[NFTA_TUNNEL_MODE]));
99 		if (priv->mode > NFT_TUNNEL_MODE_MAX)
100 			return -EOPNOTSUPP;
101 	} else {
102 		priv->mode = NFT_TUNNEL_MODE_NONE;
103 	}
104 
105 	priv->len = len;
106 	return nft_parse_register_store(ctx, tb[NFTA_TUNNEL_DREG], &priv->dreg,
107 					NULL, NFT_DATA_VALUE, len);
108 }
109 
110 static int nft_tunnel_get_dump(struct sk_buff *skb,
111 			       const struct nft_expr *expr)
112 {
113 	const struct nft_tunnel *priv = nft_expr_priv(expr);
114 
115 	if (nla_put_be32(skb, NFTA_TUNNEL_KEY, htonl(priv->key)))
116 		goto nla_put_failure;
117 	if (nft_dump_register(skb, NFTA_TUNNEL_DREG, priv->dreg))
118 		goto nla_put_failure;
119 	if (nla_put_be32(skb, NFTA_TUNNEL_MODE, htonl(priv->mode)))
120 		goto nla_put_failure;
121 	return 0;
122 
123 nla_put_failure:
124 	return -1;
125 }
126 
127 static bool nft_tunnel_get_reduce(struct nft_regs_track *track,
128 				  const struct nft_expr *expr)
129 {
130 	const struct nft_tunnel *priv = nft_expr_priv(expr);
131 	const struct nft_tunnel *tunnel;
132 
133 	if (!nft_reg_track_cmp(track, expr, priv->dreg)) {
134 		nft_reg_track_update(track, expr, priv->dreg, priv->len);
135 		return false;
136 	}
137 
138 	tunnel = nft_expr_priv(track->regs[priv->dreg].selector);
139 	if (priv->key != tunnel->key ||
140 	    priv->dreg != tunnel->dreg ||
141 	    priv->mode != tunnel->mode) {
142 		nft_reg_track_update(track, expr, priv->dreg, priv->len);
143 		return false;
144 	}
145 
146 	if (!track->regs[priv->dreg].bitwise)
147 		return true;
148 
149 	return false;
150 }
151 
152 static struct nft_expr_type nft_tunnel_type;
153 static const struct nft_expr_ops nft_tunnel_get_ops = {
154 	.type		= &nft_tunnel_type,
155 	.size		= NFT_EXPR_SIZE(sizeof(struct nft_tunnel)),
156 	.eval		= nft_tunnel_get_eval,
157 	.init		= nft_tunnel_get_init,
158 	.dump		= nft_tunnel_get_dump,
159 	.reduce		= nft_tunnel_get_reduce,
160 };
161 
162 static struct nft_expr_type nft_tunnel_type __read_mostly = {
163 	.name		= "tunnel",
164 	.ops		= &nft_tunnel_get_ops,
165 	.policy		= nft_tunnel_policy,
166 	.maxattr	= NFTA_TUNNEL_MAX,
167 	.owner		= THIS_MODULE,
168 };
169 
170 struct nft_tunnel_opts {
171 	union {
172 		struct vxlan_metadata	vxlan;
173 		struct erspan_metadata	erspan;
174 		u8	data[IP_TUNNEL_OPTS_MAX];
175 	} u;
176 	u32	len;
177 	__be16	flags;
178 };
179 
180 struct nft_tunnel_obj {
181 	struct metadata_dst	*md;
182 	struct nft_tunnel_opts	opts;
183 };
184 
185 static const struct nla_policy nft_tunnel_ip_policy[NFTA_TUNNEL_KEY_IP_MAX + 1] = {
186 	[NFTA_TUNNEL_KEY_IP_SRC]	= { .type = NLA_U32 },
187 	[NFTA_TUNNEL_KEY_IP_DST]	= { .type = NLA_U32 },
188 };
189 
190 static int nft_tunnel_obj_ip_init(const struct nft_ctx *ctx,
191 				  const struct nlattr *attr,
192 				  struct ip_tunnel_info *info)
193 {
194 	struct nlattr *tb[NFTA_TUNNEL_KEY_IP_MAX + 1];
195 	int err;
196 
197 	err = nla_parse_nested_deprecated(tb, NFTA_TUNNEL_KEY_IP_MAX, attr,
198 					  nft_tunnel_ip_policy, NULL);
199 	if (err < 0)
200 		return err;
201 
202 	if (!tb[NFTA_TUNNEL_KEY_IP_DST])
203 		return -EINVAL;
204 
205 	if (tb[NFTA_TUNNEL_KEY_IP_SRC])
206 		info->key.u.ipv4.src = nla_get_be32(tb[NFTA_TUNNEL_KEY_IP_SRC]);
207 	if (tb[NFTA_TUNNEL_KEY_IP_DST])
208 		info->key.u.ipv4.dst = nla_get_be32(tb[NFTA_TUNNEL_KEY_IP_DST]);
209 
210 	return 0;
211 }
212 
213 static const struct nla_policy nft_tunnel_ip6_policy[NFTA_TUNNEL_KEY_IP6_MAX + 1] = {
214 	[NFTA_TUNNEL_KEY_IP6_SRC]	= { .len = sizeof(struct in6_addr), },
215 	[NFTA_TUNNEL_KEY_IP6_DST]	= { .len = sizeof(struct in6_addr), },
216 	[NFTA_TUNNEL_KEY_IP6_FLOWLABEL]	= { .type = NLA_U32, }
217 };
218 
219 static int nft_tunnel_obj_ip6_init(const struct nft_ctx *ctx,
220 				   const struct nlattr *attr,
221 				   struct ip_tunnel_info *info)
222 {
223 	struct nlattr *tb[NFTA_TUNNEL_KEY_IP6_MAX + 1];
224 	int err;
225 
226 	err = nla_parse_nested_deprecated(tb, NFTA_TUNNEL_KEY_IP6_MAX, attr,
227 					  nft_tunnel_ip6_policy, NULL);
228 	if (err < 0)
229 		return err;
230 
231 	if (!tb[NFTA_TUNNEL_KEY_IP6_DST])
232 		return -EINVAL;
233 
234 	if (tb[NFTA_TUNNEL_KEY_IP6_SRC]) {
235 		memcpy(&info->key.u.ipv6.src,
236 		       nla_data(tb[NFTA_TUNNEL_KEY_IP6_SRC]),
237 		       sizeof(struct in6_addr));
238 	}
239 	if (tb[NFTA_TUNNEL_KEY_IP6_DST]) {
240 		memcpy(&info->key.u.ipv6.dst,
241 		       nla_data(tb[NFTA_TUNNEL_KEY_IP6_DST]),
242 		       sizeof(struct in6_addr));
243 	}
244 	if (tb[NFTA_TUNNEL_KEY_IP6_FLOWLABEL])
245 		info->key.label = nla_get_be32(tb[NFTA_TUNNEL_KEY_IP6_FLOWLABEL]);
246 
247 	info->mode |= IP_TUNNEL_INFO_IPV6;
248 
249 	return 0;
250 }
251 
252 static const struct nla_policy nft_tunnel_opts_vxlan_policy[NFTA_TUNNEL_KEY_VXLAN_MAX + 1] = {
253 	[NFTA_TUNNEL_KEY_VXLAN_GBP]	= { .type = NLA_U32 },
254 };
255 
256 static int nft_tunnel_obj_vxlan_init(const struct nlattr *attr,
257 				     struct nft_tunnel_opts *opts)
258 {
259 	struct nlattr *tb[NFTA_TUNNEL_KEY_VXLAN_MAX + 1];
260 	int err;
261 
262 	err = nla_parse_nested_deprecated(tb, NFTA_TUNNEL_KEY_VXLAN_MAX, attr,
263 					  nft_tunnel_opts_vxlan_policy, NULL);
264 	if (err < 0)
265 		return err;
266 
267 	if (!tb[NFTA_TUNNEL_KEY_VXLAN_GBP])
268 		return -EINVAL;
269 
270 	opts->u.vxlan.gbp = ntohl(nla_get_be32(tb[NFTA_TUNNEL_KEY_VXLAN_GBP]));
271 
272 	opts->len	= sizeof(struct vxlan_metadata);
273 	opts->flags	= TUNNEL_VXLAN_OPT;
274 
275 	return 0;
276 }
277 
278 static const struct nla_policy nft_tunnel_opts_erspan_policy[NFTA_TUNNEL_KEY_ERSPAN_MAX + 1] = {
279 	[NFTA_TUNNEL_KEY_ERSPAN_VERSION]	= { .type = NLA_U32 },
280 	[NFTA_TUNNEL_KEY_ERSPAN_V1_INDEX]	= { .type = NLA_U32 },
281 	[NFTA_TUNNEL_KEY_ERSPAN_V2_DIR]		= { .type = NLA_U8 },
282 	[NFTA_TUNNEL_KEY_ERSPAN_V2_HWID]	= { .type = NLA_U8 },
283 };
284 
285 static int nft_tunnel_obj_erspan_init(const struct nlattr *attr,
286 				      struct nft_tunnel_opts *opts)
287 {
288 	struct nlattr *tb[NFTA_TUNNEL_KEY_ERSPAN_MAX + 1];
289 	uint8_t hwid, dir;
290 	int err, version;
291 
292 	err = nla_parse_nested_deprecated(tb, NFTA_TUNNEL_KEY_ERSPAN_MAX,
293 					  attr, nft_tunnel_opts_erspan_policy,
294 					  NULL);
295 	if (err < 0)
296 		return err;
297 
298 	if (!tb[NFTA_TUNNEL_KEY_ERSPAN_VERSION])
299 		 return -EINVAL;
300 
301 	version = ntohl(nla_get_be32(tb[NFTA_TUNNEL_KEY_ERSPAN_VERSION]));
302 	switch (version) {
303 	case ERSPAN_VERSION:
304 		if (!tb[NFTA_TUNNEL_KEY_ERSPAN_V1_INDEX])
305 			return -EINVAL;
306 
307 		opts->u.erspan.u.index =
308 			nla_get_be32(tb[NFTA_TUNNEL_KEY_ERSPAN_V1_INDEX]);
309 		break;
310 	case ERSPAN_VERSION2:
311 		if (!tb[NFTA_TUNNEL_KEY_ERSPAN_V2_DIR] ||
312 		    !tb[NFTA_TUNNEL_KEY_ERSPAN_V2_HWID])
313 			return -EINVAL;
314 
315 		hwid = nla_get_u8(tb[NFTA_TUNNEL_KEY_ERSPAN_V2_HWID]);
316 		dir = nla_get_u8(tb[NFTA_TUNNEL_KEY_ERSPAN_V2_DIR]);
317 
318 		set_hwid(&opts->u.erspan.u.md2, hwid);
319 		opts->u.erspan.u.md2.dir = dir;
320 		break;
321 	default:
322 		return -EOPNOTSUPP;
323 	}
324 	opts->u.erspan.version = version;
325 
326 	opts->len	= sizeof(struct erspan_metadata);
327 	opts->flags	= TUNNEL_ERSPAN_OPT;
328 
329 	return 0;
330 }
331 
332 static const struct nla_policy nft_tunnel_opts_geneve_policy[NFTA_TUNNEL_KEY_GENEVE_MAX + 1] = {
333 	[NFTA_TUNNEL_KEY_GENEVE_CLASS]	= { .type = NLA_U16 },
334 	[NFTA_TUNNEL_KEY_GENEVE_TYPE]	= { .type = NLA_U8 },
335 	[NFTA_TUNNEL_KEY_GENEVE_DATA]	= { .type = NLA_BINARY, .len = 128 },
336 };
337 
338 static int nft_tunnel_obj_geneve_init(const struct nlattr *attr,
339 				      struct nft_tunnel_opts *opts)
340 {
341 	struct geneve_opt *opt = (struct geneve_opt *)opts->u.data + opts->len;
342 	struct nlattr *tb[NFTA_TUNNEL_KEY_GENEVE_MAX + 1];
343 	int err, data_len;
344 
345 	err = nla_parse_nested(tb, NFTA_TUNNEL_KEY_GENEVE_MAX, attr,
346 			       nft_tunnel_opts_geneve_policy, NULL);
347 	if (err < 0)
348 		return err;
349 
350 	if (!tb[NFTA_TUNNEL_KEY_GENEVE_CLASS] ||
351 	    !tb[NFTA_TUNNEL_KEY_GENEVE_TYPE] ||
352 	    !tb[NFTA_TUNNEL_KEY_GENEVE_DATA])
353 		return -EINVAL;
354 
355 	attr = tb[NFTA_TUNNEL_KEY_GENEVE_DATA];
356 	data_len = nla_len(attr);
357 	if (data_len % 4)
358 		return -EINVAL;
359 
360 	opts->len += sizeof(*opt) + data_len;
361 	if (opts->len > IP_TUNNEL_OPTS_MAX)
362 		return -EINVAL;
363 
364 	memcpy(opt->opt_data, nla_data(attr), data_len);
365 	opt->length = data_len / 4;
366 	opt->opt_class = nla_get_be16(tb[NFTA_TUNNEL_KEY_GENEVE_CLASS]);
367 	opt->type = nla_get_u8(tb[NFTA_TUNNEL_KEY_GENEVE_TYPE]);
368 	opts->flags = TUNNEL_GENEVE_OPT;
369 
370 	return 0;
371 }
372 
373 static const struct nla_policy nft_tunnel_opts_policy[NFTA_TUNNEL_KEY_OPTS_MAX + 1] = {
374 	[NFTA_TUNNEL_KEY_OPTS_UNSPEC]	= {
375 		.strict_start_type = NFTA_TUNNEL_KEY_OPTS_GENEVE },
376 	[NFTA_TUNNEL_KEY_OPTS_VXLAN]	= { .type = NLA_NESTED, },
377 	[NFTA_TUNNEL_KEY_OPTS_ERSPAN]	= { .type = NLA_NESTED, },
378 	[NFTA_TUNNEL_KEY_OPTS_GENEVE]	= { .type = NLA_NESTED, },
379 };
380 
381 static int nft_tunnel_obj_opts_init(const struct nft_ctx *ctx,
382 				    const struct nlattr *attr,
383 				    struct ip_tunnel_info *info,
384 				    struct nft_tunnel_opts *opts)
385 {
386 	int err, rem, type = 0;
387 	struct nlattr *nla;
388 
389 	err = nla_validate_nested_deprecated(attr, NFTA_TUNNEL_KEY_OPTS_MAX,
390 					     nft_tunnel_opts_policy, NULL);
391 	if (err < 0)
392 		return err;
393 
394 	nla_for_each_attr(nla, nla_data(attr), nla_len(attr), rem) {
395 		switch (nla_type(nla)) {
396 		case NFTA_TUNNEL_KEY_OPTS_VXLAN:
397 			if (type)
398 				return -EINVAL;
399 			err = nft_tunnel_obj_vxlan_init(nla, opts);
400 			if (err)
401 				return err;
402 			type = TUNNEL_VXLAN_OPT;
403 			break;
404 		case NFTA_TUNNEL_KEY_OPTS_ERSPAN:
405 			if (type)
406 				return -EINVAL;
407 			err = nft_tunnel_obj_erspan_init(nla, opts);
408 			if (err)
409 				return err;
410 			type = TUNNEL_ERSPAN_OPT;
411 			break;
412 		case NFTA_TUNNEL_KEY_OPTS_GENEVE:
413 			if (type && type != TUNNEL_GENEVE_OPT)
414 				return -EINVAL;
415 			err = nft_tunnel_obj_geneve_init(nla, opts);
416 			if (err)
417 				return err;
418 			type = TUNNEL_GENEVE_OPT;
419 			break;
420 		default:
421 			return -EOPNOTSUPP;
422 		}
423 	}
424 
425 	return err;
426 }
427 
428 static const struct nla_policy nft_tunnel_key_policy[NFTA_TUNNEL_KEY_MAX + 1] = {
429 	[NFTA_TUNNEL_KEY_IP]	= { .type = NLA_NESTED, },
430 	[NFTA_TUNNEL_KEY_IP6]	= { .type = NLA_NESTED, },
431 	[NFTA_TUNNEL_KEY_ID]	= { .type = NLA_U32, },
432 	[NFTA_TUNNEL_KEY_FLAGS]	= { .type = NLA_U32, },
433 	[NFTA_TUNNEL_KEY_TOS]	= { .type = NLA_U8, },
434 	[NFTA_TUNNEL_KEY_TTL]	= { .type = NLA_U8, },
435 	[NFTA_TUNNEL_KEY_SPORT]	= { .type = NLA_U16, },
436 	[NFTA_TUNNEL_KEY_DPORT]	= { .type = NLA_U16, },
437 	[NFTA_TUNNEL_KEY_OPTS]	= { .type = NLA_NESTED, },
438 };
439 
440 static int nft_tunnel_obj_init(const struct nft_ctx *ctx,
441 			       const struct nlattr * const tb[],
442 			       struct nft_object *obj)
443 {
444 	struct nft_tunnel_obj *priv = nft_obj_data(obj);
445 	struct ip_tunnel_info info;
446 	struct metadata_dst *md;
447 	int err;
448 
449 	if (!tb[NFTA_TUNNEL_KEY_ID])
450 		return -EINVAL;
451 
452 	memset(&info, 0, sizeof(info));
453 	info.mode		= IP_TUNNEL_INFO_TX;
454 	info.key.tun_id		= key32_to_tunnel_id(nla_get_be32(tb[NFTA_TUNNEL_KEY_ID]));
455 	info.key.tun_flags	= TUNNEL_KEY | TUNNEL_CSUM | TUNNEL_NOCACHE;
456 
457 	if (tb[NFTA_TUNNEL_KEY_IP]) {
458 		err = nft_tunnel_obj_ip_init(ctx, tb[NFTA_TUNNEL_KEY_IP], &info);
459 		if (err < 0)
460 			return err;
461 	} else if (tb[NFTA_TUNNEL_KEY_IP6]) {
462 		err = nft_tunnel_obj_ip6_init(ctx, tb[NFTA_TUNNEL_KEY_IP6], &info);
463 		if (err < 0)
464 			return err;
465 	} else {
466 		return -EINVAL;
467 	}
468 
469 	if (tb[NFTA_TUNNEL_KEY_SPORT]) {
470 		info.key.tp_src = nla_get_be16(tb[NFTA_TUNNEL_KEY_SPORT]);
471 	}
472 	if (tb[NFTA_TUNNEL_KEY_DPORT]) {
473 		info.key.tp_dst = nla_get_be16(tb[NFTA_TUNNEL_KEY_DPORT]);
474 	}
475 
476 	if (tb[NFTA_TUNNEL_KEY_FLAGS]) {
477 		u32 tun_flags;
478 
479 		tun_flags = ntohl(nla_get_be32(tb[NFTA_TUNNEL_KEY_FLAGS]));
480 		if (tun_flags & ~NFT_TUNNEL_F_MASK)
481 			return -EOPNOTSUPP;
482 
483 		if (tun_flags & NFT_TUNNEL_F_ZERO_CSUM_TX)
484 			info.key.tun_flags &= ~TUNNEL_CSUM;
485 		if (tun_flags & NFT_TUNNEL_F_DONT_FRAGMENT)
486 			info.key.tun_flags |= TUNNEL_DONT_FRAGMENT;
487 		if (tun_flags & NFT_TUNNEL_F_SEQ_NUMBER)
488 			info.key.tun_flags |= TUNNEL_SEQ;
489 	}
490 	if (tb[NFTA_TUNNEL_KEY_TOS])
491 		info.key.tos = nla_get_u8(tb[NFTA_TUNNEL_KEY_TOS]);
492 	if (tb[NFTA_TUNNEL_KEY_TTL])
493 		info.key.ttl = nla_get_u8(tb[NFTA_TUNNEL_KEY_TTL]);
494 	else
495 		info.key.ttl = U8_MAX;
496 
497 	if (tb[NFTA_TUNNEL_KEY_OPTS]) {
498 		err = nft_tunnel_obj_opts_init(ctx, tb[NFTA_TUNNEL_KEY_OPTS],
499 					       &info, &priv->opts);
500 		if (err < 0)
501 			return err;
502 	}
503 
504 	md = metadata_dst_alloc(priv->opts.len, METADATA_IP_TUNNEL, GFP_KERNEL);
505 	if (!md)
506 		return -ENOMEM;
507 
508 	memcpy(&md->u.tun_info, &info, sizeof(info));
509 #ifdef CONFIG_DST_CACHE
510 	err = dst_cache_init(&md->u.tun_info.dst_cache, GFP_KERNEL);
511 	if (err < 0) {
512 		metadata_dst_free(md);
513 		return err;
514 	}
515 #endif
516 	ip_tunnel_info_opts_set(&md->u.tun_info, &priv->opts.u, priv->opts.len,
517 				priv->opts.flags);
518 	priv->md = md;
519 
520 	return 0;
521 }
522 
523 static inline void nft_tunnel_obj_eval(struct nft_object *obj,
524 				       struct nft_regs *regs,
525 				       const struct nft_pktinfo *pkt)
526 {
527 	struct nft_tunnel_obj *priv = nft_obj_data(obj);
528 	struct sk_buff *skb = pkt->skb;
529 
530 	skb_dst_drop(skb);
531 	dst_hold((struct dst_entry *) priv->md);
532 	skb_dst_set(skb, (struct dst_entry *) priv->md);
533 }
534 
535 static int nft_tunnel_ip_dump(struct sk_buff *skb, struct ip_tunnel_info *info)
536 {
537 	struct nlattr *nest;
538 
539 	if (info->mode & IP_TUNNEL_INFO_IPV6) {
540 		nest = nla_nest_start_noflag(skb, NFTA_TUNNEL_KEY_IP6);
541 		if (!nest)
542 			return -1;
543 
544 		if (nla_put_in6_addr(skb, NFTA_TUNNEL_KEY_IP6_SRC,
545 				     &info->key.u.ipv6.src) < 0 ||
546 		    nla_put_in6_addr(skb, NFTA_TUNNEL_KEY_IP6_DST,
547 				     &info->key.u.ipv6.dst) < 0 ||
548 		    nla_put_be32(skb, NFTA_TUNNEL_KEY_IP6_FLOWLABEL,
549 				 info->key.label)) {
550 			nla_nest_cancel(skb, nest);
551 			return -1;
552 		}
553 
554 		nla_nest_end(skb, nest);
555 	} else {
556 		nest = nla_nest_start_noflag(skb, NFTA_TUNNEL_KEY_IP);
557 		if (!nest)
558 			return -1;
559 
560 		if (nla_put_in_addr(skb, NFTA_TUNNEL_KEY_IP_SRC,
561 				    info->key.u.ipv4.src) < 0 ||
562 		    nla_put_in_addr(skb, NFTA_TUNNEL_KEY_IP_DST,
563 				    info->key.u.ipv4.dst) < 0) {
564 			nla_nest_cancel(skb, nest);
565 			return -1;
566 		}
567 
568 		nla_nest_end(skb, nest);
569 	}
570 
571 	return 0;
572 }
573 
574 static int nft_tunnel_opts_dump(struct sk_buff *skb,
575 				struct nft_tunnel_obj *priv)
576 {
577 	struct nft_tunnel_opts *opts = &priv->opts;
578 	struct nlattr *nest, *inner;
579 
580 	nest = nla_nest_start_noflag(skb, NFTA_TUNNEL_KEY_OPTS);
581 	if (!nest)
582 		return -1;
583 
584 	if (opts->flags & TUNNEL_VXLAN_OPT) {
585 		inner = nla_nest_start_noflag(skb, NFTA_TUNNEL_KEY_OPTS_VXLAN);
586 		if (!inner)
587 			goto failure;
588 		if (nla_put_be32(skb, NFTA_TUNNEL_KEY_VXLAN_GBP,
589 				 htonl(opts->u.vxlan.gbp)))
590 			goto inner_failure;
591 		nla_nest_end(skb, inner);
592 	} else if (opts->flags & TUNNEL_ERSPAN_OPT) {
593 		inner = nla_nest_start_noflag(skb, NFTA_TUNNEL_KEY_OPTS_ERSPAN);
594 		if (!inner)
595 			goto failure;
596 		if (nla_put_be32(skb, NFTA_TUNNEL_KEY_ERSPAN_VERSION,
597 				 htonl(opts->u.erspan.version)))
598 			goto inner_failure;
599 		switch (opts->u.erspan.version) {
600 		case ERSPAN_VERSION:
601 			if (nla_put_be32(skb, NFTA_TUNNEL_KEY_ERSPAN_V1_INDEX,
602 					 opts->u.erspan.u.index))
603 				goto inner_failure;
604 			break;
605 		case ERSPAN_VERSION2:
606 			if (nla_put_u8(skb, NFTA_TUNNEL_KEY_ERSPAN_V2_HWID,
607 				       get_hwid(&opts->u.erspan.u.md2)) ||
608 			    nla_put_u8(skb, NFTA_TUNNEL_KEY_ERSPAN_V2_DIR,
609 				       opts->u.erspan.u.md2.dir))
610 				goto inner_failure;
611 			break;
612 		}
613 		nla_nest_end(skb, inner);
614 	} else if (opts->flags & TUNNEL_GENEVE_OPT) {
615 		struct geneve_opt *opt;
616 		int offset = 0;
617 
618 		inner = nla_nest_start_noflag(skb, NFTA_TUNNEL_KEY_OPTS_GENEVE);
619 		if (!inner)
620 			goto failure;
621 		while (opts->len > offset) {
622 			opt = (struct geneve_opt *)opts->u.data + offset;
623 			if (nla_put_be16(skb, NFTA_TUNNEL_KEY_GENEVE_CLASS,
624 					 opt->opt_class) ||
625 			    nla_put_u8(skb, NFTA_TUNNEL_KEY_GENEVE_TYPE,
626 				       opt->type) ||
627 			    nla_put(skb, NFTA_TUNNEL_KEY_GENEVE_DATA,
628 				    opt->length * 4, opt->opt_data))
629 				goto inner_failure;
630 			offset += sizeof(*opt) + opt->length * 4;
631 		}
632 		nla_nest_end(skb, inner);
633 	}
634 	nla_nest_end(skb, nest);
635 	return 0;
636 
637 inner_failure:
638 	nla_nest_cancel(skb, inner);
639 failure:
640 	nla_nest_cancel(skb, nest);
641 	return -1;
642 }
643 
644 static int nft_tunnel_ports_dump(struct sk_buff *skb,
645 				 struct ip_tunnel_info *info)
646 {
647 	if (nla_put_be16(skb, NFTA_TUNNEL_KEY_SPORT, info->key.tp_src) < 0 ||
648 	    nla_put_be16(skb, NFTA_TUNNEL_KEY_DPORT, info->key.tp_dst) < 0)
649 		return -1;
650 
651 	return 0;
652 }
653 
654 static int nft_tunnel_flags_dump(struct sk_buff *skb,
655 				 struct ip_tunnel_info *info)
656 {
657 	u32 flags = 0;
658 
659 	if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT)
660 		flags |= NFT_TUNNEL_F_DONT_FRAGMENT;
661 	if (!(info->key.tun_flags & TUNNEL_CSUM))
662 		flags |= NFT_TUNNEL_F_ZERO_CSUM_TX;
663 	if (info->key.tun_flags & TUNNEL_SEQ)
664 		flags |= NFT_TUNNEL_F_SEQ_NUMBER;
665 
666 	if (nla_put_be32(skb, NFTA_TUNNEL_KEY_FLAGS, htonl(flags)) < 0)
667 		return -1;
668 
669 	return 0;
670 }
671 
672 static int nft_tunnel_obj_dump(struct sk_buff *skb,
673 			       struct nft_object *obj, bool reset)
674 {
675 	struct nft_tunnel_obj *priv = nft_obj_data(obj);
676 	struct ip_tunnel_info *info = &priv->md->u.tun_info;
677 
678 	if (nla_put_be32(skb, NFTA_TUNNEL_KEY_ID,
679 			 tunnel_id_to_key32(info->key.tun_id)) ||
680 	    nft_tunnel_ip_dump(skb, info) < 0 ||
681 	    nft_tunnel_ports_dump(skb, info) < 0 ||
682 	    nft_tunnel_flags_dump(skb, info) < 0 ||
683 	    nla_put_u8(skb, NFTA_TUNNEL_KEY_TOS, info->key.tos) ||
684 	    nla_put_u8(skb, NFTA_TUNNEL_KEY_TTL, info->key.ttl) ||
685 	    nft_tunnel_opts_dump(skb, priv) < 0)
686 		goto nla_put_failure;
687 
688 	return 0;
689 
690 nla_put_failure:
691 	return -1;
692 }
693 
694 static void nft_tunnel_obj_destroy(const struct nft_ctx *ctx,
695 				   struct nft_object *obj)
696 {
697 	struct nft_tunnel_obj *priv = nft_obj_data(obj);
698 
699 	metadata_dst_free(priv->md);
700 }
701 
702 static struct nft_object_type nft_tunnel_obj_type;
703 static const struct nft_object_ops nft_tunnel_obj_ops = {
704 	.type		= &nft_tunnel_obj_type,
705 	.size		= sizeof(struct nft_tunnel_obj),
706 	.eval		= nft_tunnel_obj_eval,
707 	.init		= nft_tunnel_obj_init,
708 	.destroy	= nft_tunnel_obj_destroy,
709 	.dump		= nft_tunnel_obj_dump,
710 };
711 
712 static struct nft_object_type nft_tunnel_obj_type __read_mostly = {
713 	.type		= NFT_OBJECT_TUNNEL,
714 	.ops		= &nft_tunnel_obj_ops,
715 	.maxattr	= NFTA_TUNNEL_KEY_MAX,
716 	.policy		= nft_tunnel_key_policy,
717 	.owner		= THIS_MODULE,
718 };
719 
720 static int __init nft_tunnel_module_init(void)
721 {
722 	int err;
723 
724 	err = nft_register_expr(&nft_tunnel_type);
725 	if (err < 0)
726 		return err;
727 
728 	err = nft_register_obj(&nft_tunnel_obj_type);
729 	if (err < 0)
730 		nft_unregister_expr(&nft_tunnel_type);
731 
732 	return err;
733 }
734 
735 static void __exit nft_tunnel_module_exit(void)
736 {
737 	nft_unregister_obj(&nft_tunnel_obj_type);
738 	nft_unregister_expr(&nft_tunnel_type);
739 }
740 
741 module_init(nft_tunnel_module_init);
742 module_exit(nft_tunnel_module_exit);
743 
744 MODULE_LICENSE("GPL");
745 MODULE_AUTHOR("Pablo Neira Ayuso <pablo@netfilter.org>");
746 MODULE_ALIAS_NFT_EXPR("tunnel");
747 MODULE_ALIAS_NFT_OBJ(NFT_OBJECT_TUNNEL);
748 MODULE_DESCRIPTION("nftables tunnel expression support");
749