xref: /linux/net/sched/act_tunnel_key.c (revision 48dea9a700c8728cc31a1dd44588b97578de86ee)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  * Copyright (c) 2016, Amir Vadai <amir@vadai.me>
4  * Copyright (c) 2016, Mellanox Technologies. All rights reserved.
5  */
6 
7 #include <linux/module.h>
8 #include <linux/init.h>
9 #include <linux/kernel.h>
10 #include <linux/skbuff.h>
11 #include <linux/rtnetlink.h>
12 #include <net/geneve.h>
13 #include <net/vxlan.h>
14 #include <net/erspan.h>
15 #include <net/netlink.h>
16 #include <net/pkt_sched.h>
17 #include <net/dst.h>
18 #include <net/pkt_cls.h>
19 
20 #include <linux/tc_act/tc_tunnel_key.h>
21 #include <net/tc_act/tc_tunnel_key.h>
22 
23 static unsigned int tunnel_key_net_id;
24 static struct tc_action_ops act_tunnel_key_ops;
25 
26 static int tunnel_key_act(struct sk_buff *skb, const struct tc_action *a,
27 			  struct tcf_result *res)
28 {
29 	struct tcf_tunnel_key *t = to_tunnel_key(a);
30 	struct tcf_tunnel_key_params *params;
31 	int action;
32 
33 	params = rcu_dereference_bh(t->params);
34 
35 	tcf_lastuse_update(&t->tcf_tm);
36 	tcf_action_update_bstats(&t->common, skb);
37 	action = READ_ONCE(t->tcf_action);
38 
39 	switch (params->tcft_action) {
40 	case TCA_TUNNEL_KEY_ACT_RELEASE:
41 		skb_dst_drop(skb);
42 		break;
43 	case TCA_TUNNEL_KEY_ACT_SET:
44 		skb_dst_drop(skb);
45 		skb_dst_set(skb, dst_clone(&params->tcft_enc_metadata->dst));
46 		break;
47 	default:
48 		WARN_ONCE(1, "Bad tunnel_key action %d.\n",
49 			  params->tcft_action);
50 		break;
51 	}
52 
53 	return action;
54 }
55 
56 static const struct nla_policy
57 enc_opts_policy[TCA_TUNNEL_KEY_ENC_OPTS_MAX + 1] = {
58 	[TCA_TUNNEL_KEY_ENC_OPTS_UNSPEC]	= {
59 		.strict_start_type = TCA_TUNNEL_KEY_ENC_OPTS_VXLAN },
60 	[TCA_TUNNEL_KEY_ENC_OPTS_GENEVE]	= { .type = NLA_NESTED },
61 	[TCA_TUNNEL_KEY_ENC_OPTS_VXLAN]		= { .type = NLA_NESTED },
62 	[TCA_TUNNEL_KEY_ENC_OPTS_ERSPAN]	= { .type = NLA_NESTED },
63 };
64 
65 static const struct nla_policy
66 geneve_opt_policy[TCA_TUNNEL_KEY_ENC_OPT_GENEVE_MAX + 1] = {
67 	[TCA_TUNNEL_KEY_ENC_OPT_GENEVE_CLASS]	   = { .type = NLA_U16 },
68 	[TCA_TUNNEL_KEY_ENC_OPT_GENEVE_TYPE]	   = { .type = NLA_U8 },
69 	[TCA_TUNNEL_KEY_ENC_OPT_GENEVE_DATA]	   = { .type = NLA_BINARY,
70 						       .len = 128 },
71 };
72 
73 static const struct nla_policy
74 vxlan_opt_policy[TCA_TUNNEL_KEY_ENC_OPT_VXLAN_MAX + 1] = {
75 	[TCA_TUNNEL_KEY_ENC_OPT_VXLAN_GBP]	   = { .type = NLA_U32 },
76 };
77 
78 static const struct nla_policy
79 erspan_opt_policy[TCA_TUNNEL_KEY_ENC_OPT_ERSPAN_MAX + 1] = {
80 	[TCA_TUNNEL_KEY_ENC_OPT_ERSPAN_VER]	   = { .type = NLA_U8 },
81 	[TCA_TUNNEL_KEY_ENC_OPT_ERSPAN_INDEX]	   = { .type = NLA_U32 },
82 	[TCA_TUNNEL_KEY_ENC_OPT_ERSPAN_DIR]	   = { .type = NLA_U8 },
83 	[TCA_TUNNEL_KEY_ENC_OPT_ERSPAN_HWID]	   = { .type = NLA_U8 },
84 };
85 
86 static int
87 tunnel_key_copy_geneve_opt(const struct nlattr *nla, void *dst, int dst_len,
88 			   struct netlink_ext_ack *extack)
89 {
90 	struct nlattr *tb[TCA_TUNNEL_KEY_ENC_OPT_GENEVE_MAX + 1];
91 	int err, data_len, opt_len;
92 	u8 *data;
93 
94 	err = nla_parse_nested_deprecated(tb,
95 					  TCA_TUNNEL_KEY_ENC_OPT_GENEVE_MAX,
96 					  nla, geneve_opt_policy, extack);
97 	if (err < 0)
98 		return err;
99 
100 	if (!tb[TCA_TUNNEL_KEY_ENC_OPT_GENEVE_CLASS] ||
101 	    !tb[TCA_TUNNEL_KEY_ENC_OPT_GENEVE_TYPE] ||
102 	    !tb[TCA_TUNNEL_KEY_ENC_OPT_GENEVE_DATA]) {
103 		NL_SET_ERR_MSG(extack, "Missing tunnel key geneve option class, type or data");
104 		return -EINVAL;
105 	}
106 
107 	data = nla_data(tb[TCA_TUNNEL_KEY_ENC_OPT_GENEVE_DATA]);
108 	data_len = nla_len(tb[TCA_TUNNEL_KEY_ENC_OPT_GENEVE_DATA]);
109 	if (data_len < 4) {
110 		NL_SET_ERR_MSG(extack, "Tunnel key geneve option data is less than 4 bytes long");
111 		return -ERANGE;
112 	}
113 	if (data_len % 4) {
114 		NL_SET_ERR_MSG(extack, "Tunnel key geneve option data is not a multiple of 4 bytes long");
115 		return -ERANGE;
116 	}
117 
118 	opt_len = sizeof(struct geneve_opt) + data_len;
119 	if (dst) {
120 		struct geneve_opt *opt = dst;
121 
122 		WARN_ON(dst_len < opt_len);
123 
124 		opt->opt_class =
125 			nla_get_be16(tb[TCA_TUNNEL_KEY_ENC_OPT_GENEVE_CLASS]);
126 		opt->type = nla_get_u8(tb[TCA_TUNNEL_KEY_ENC_OPT_GENEVE_TYPE]);
127 		opt->length = data_len / 4; /* length is in units of 4 bytes */
128 		opt->r1 = 0;
129 		opt->r2 = 0;
130 		opt->r3 = 0;
131 
132 		memcpy(opt + 1, data, data_len);
133 	}
134 
135 	return opt_len;
136 }
137 
138 static int
139 tunnel_key_copy_vxlan_opt(const struct nlattr *nla, void *dst, int dst_len,
140 			  struct netlink_ext_ack *extack)
141 {
142 	struct nlattr *tb[TCA_TUNNEL_KEY_ENC_OPT_VXLAN_MAX + 1];
143 	int err;
144 
145 	err = nla_parse_nested(tb, TCA_TUNNEL_KEY_ENC_OPT_VXLAN_MAX, nla,
146 			       vxlan_opt_policy, extack);
147 	if (err < 0)
148 		return err;
149 
150 	if (!tb[TCA_TUNNEL_KEY_ENC_OPT_VXLAN_GBP]) {
151 		NL_SET_ERR_MSG(extack, "Missing tunnel key vxlan option gbp");
152 		return -EINVAL;
153 	}
154 
155 	if (dst) {
156 		struct vxlan_metadata *md = dst;
157 
158 		md->gbp = nla_get_u32(tb[TCA_TUNNEL_KEY_ENC_OPT_VXLAN_GBP]);
159 	}
160 
161 	return sizeof(struct vxlan_metadata);
162 }
163 
164 static int
165 tunnel_key_copy_erspan_opt(const struct nlattr *nla, void *dst, int dst_len,
166 			   struct netlink_ext_ack *extack)
167 {
168 	struct nlattr *tb[TCA_TUNNEL_KEY_ENC_OPT_ERSPAN_MAX + 1];
169 	int err;
170 	u8 ver;
171 
172 	err = nla_parse_nested(tb, TCA_TUNNEL_KEY_ENC_OPT_ERSPAN_MAX, nla,
173 			       erspan_opt_policy, extack);
174 	if (err < 0)
175 		return err;
176 
177 	if (!tb[TCA_TUNNEL_KEY_ENC_OPT_ERSPAN_VER]) {
178 		NL_SET_ERR_MSG(extack, "Missing tunnel key erspan option ver");
179 		return -EINVAL;
180 	}
181 
182 	ver = nla_get_u8(tb[TCA_TUNNEL_KEY_ENC_OPT_ERSPAN_VER]);
183 	if (ver == 1) {
184 		if (!tb[TCA_TUNNEL_KEY_ENC_OPT_ERSPAN_INDEX]) {
185 			NL_SET_ERR_MSG(extack, "Missing tunnel key erspan option index");
186 			return -EINVAL;
187 		}
188 	} else if (ver == 2) {
189 		if (!tb[TCA_TUNNEL_KEY_ENC_OPT_ERSPAN_DIR] ||
190 		    !tb[TCA_TUNNEL_KEY_ENC_OPT_ERSPAN_HWID]) {
191 			NL_SET_ERR_MSG(extack, "Missing tunnel key erspan option dir or hwid");
192 			return -EINVAL;
193 		}
194 	} else {
195 		NL_SET_ERR_MSG(extack, "Tunnel key erspan option ver is incorrect");
196 		return -EINVAL;
197 	}
198 
199 	if (dst) {
200 		struct erspan_metadata *md = dst;
201 
202 		md->version = ver;
203 		if (ver == 1) {
204 			nla = tb[TCA_TUNNEL_KEY_ENC_OPT_ERSPAN_INDEX];
205 			md->u.index = nla_get_be32(nla);
206 		} else {
207 			nla = tb[TCA_TUNNEL_KEY_ENC_OPT_ERSPAN_DIR];
208 			md->u.md2.dir = nla_get_u8(nla);
209 			nla = tb[TCA_TUNNEL_KEY_ENC_OPT_ERSPAN_HWID];
210 			set_hwid(&md->u.md2, nla_get_u8(nla));
211 		}
212 	}
213 
214 	return sizeof(struct erspan_metadata);
215 }
216 
217 static int tunnel_key_copy_opts(const struct nlattr *nla, u8 *dst,
218 				int dst_len, struct netlink_ext_ack *extack)
219 {
220 	int err, rem, opt_len, len = nla_len(nla), opts_len = 0, type = 0;
221 	const struct nlattr *attr, *head = nla_data(nla);
222 
223 	err = nla_validate_deprecated(head, len, TCA_TUNNEL_KEY_ENC_OPTS_MAX,
224 				      enc_opts_policy, extack);
225 	if (err)
226 		return err;
227 
228 	nla_for_each_attr(attr, head, len, rem) {
229 		switch (nla_type(attr)) {
230 		case TCA_TUNNEL_KEY_ENC_OPTS_GENEVE:
231 			if (type && type != TUNNEL_GENEVE_OPT) {
232 				NL_SET_ERR_MSG(extack, "Duplicate type for geneve options");
233 				return -EINVAL;
234 			}
235 			opt_len = tunnel_key_copy_geneve_opt(attr, dst,
236 							     dst_len, extack);
237 			if (opt_len < 0)
238 				return opt_len;
239 			opts_len += opt_len;
240 			if (opts_len > IP_TUNNEL_OPTS_MAX) {
241 				NL_SET_ERR_MSG(extack, "Tunnel options exceeds max size");
242 				return -EINVAL;
243 			}
244 			if (dst) {
245 				dst_len -= opt_len;
246 				dst += opt_len;
247 			}
248 			type = TUNNEL_GENEVE_OPT;
249 			break;
250 		case TCA_TUNNEL_KEY_ENC_OPTS_VXLAN:
251 			if (type) {
252 				NL_SET_ERR_MSG(extack, "Duplicate type for vxlan options");
253 				return -EINVAL;
254 			}
255 			opt_len = tunnel_key_copy_vxlan_opt(attr, dst,
256 							    dst_len, extack);
257 			if (opt_len < 0)
258 				return opt_len;
259 			opts_len += opt_len;
260 			type = TUNNEL_VXLAN_OPT;
261 			break;
262 		case TCA_TUNNEL_KEY_ENC_OPTS_ERSPAN:
263 			if (type) {
264 				NL_SET_ERR_MSG(extack, "Duplicate type for erspan options");
265 				return -EINVAL;
266 			}
267 			opt_len = tunnel_key_copy_erspan_opt(attr, dst,
268 							     dst_len, extack);
269 			if (opt_len < 0)
270 				return opt_len;
271 			opts_len += opt_len;
272 			type = TUNNEL_ERSPAN_OPT;
273 			break;
274 		}
275 	}
276 
277 	if (!opts_len) {
278 		NL_SET_ERR_MSG(extack, "Empty list of tunnel options");
279 		return -EINVAL;
280 	}
281 
282 	if (rem > 0) {
283 		NL_SET_ERR_MSG(extack, "Trailing data after parsing tunnel key options attributes");
284 		return -EINVAL;
285 	}
286 
287 	return opts_len;
288 }
289 
290 static int tunnel_key_get_opts_len(struct nlattr *nla,
291 				   struct netlink_ext_ack *extack)
292 {
293 	return tunnel_key_copy_opts(nla, NULL, 0, extack);
294 }
295 
296 static int tunnel_key_opts_set(struct nlattr *nla, struct ip_tunnel_info *info,
297 			       int opts_len, struct netlink_ext_ack *extack)
298 {
299 	info->options_len = opts_len;
300 	switch (nla_type(nla_data(nla))) {
301 	case TCA_TUNNEL_KEY_ENC_OPTS_GENEVE:
302 #if IS_ENABLED(CONFIG_INET)
303 		info->key.tun_flags |= TUNNEL_GENEVE_OPT;
304 		return tunnel_key_copy_opts(nla, ip_tunnel_info_opts(info),
305 					    opts_len, extack);
306 #else
307 		return -EAFNOSUPPORT;
308 #endif
309 	case TCA_TUNNEL_KEY_ENC_OPTS_VXLAN:
310 #if IS_ENABLED(CONFIG_INET)
311 		info->key.tun_flags |= TUNNEL_VXLAN_OPT;
312 		return tunnel_key_copy_opts(nla, ip_tunnel_info_opts(info),
313 					    opts_len, extack);
314 #else
315 		return -EAFNOSUPPORT;
316 #endif
317 	case TCA_TUNNEL_KEY_ENC_OPTS_ERSPAN:
318 #if IS_ENABLED(CONFIG_INET)
319 		info->key.tun_flags |= TUNNEL_ERSPAN_OPT;
320 		return tunnel_key_copy_opts(nla, ip_tunnel_info_opts(info),
321 					    opts_len, extack);
322 #else
323 		return -EAFNOSUPPORT;
324 #endif
325 	default:
326 		NL_SET_ERR_MSG(extack, "Cannot set tunnel options for unknown tunnel type");
327 		return -EINVAL;
328 	}
329 }
330 
331 static const struct nla_policy tunnel_key_policy[TCA_TUNNEL_KEY_MAX + 1] = {
332 	[TCA_TUNNEL_KEY_PARMS]	    = { .len = sizeof(struct tc_tunnel_key) },
333 	[TCA_TUNNEL_KEY_ENC_IPV4_SRC] = { .type = NLA_U32 },
334 	[TCA_TUNNEL_KEY_ENC_IPV4_DST] = { .type = NLA_U32 },
335 	[TCA_TUNNEL_KEY_ENC_IPV6_SRC] = { .len = sizeof(struct in6_addr) },
336 	[TCA_TUNNEL_KEY_ENC_IPV6_DST] = { .len = sizeof(struct in6_addr) },
337 	[TCA_TUNNEL_KEY_ENC_KEY_ID]   = { .type = NLA_U32 },
338 	[TCA_TUNNEL_KEY_ENC_DST_PORT] = {.type = NLA_U16},
339 	[TCA_TUNNEL_KEY_NO_CSUM]      = { .type = NLA_U8 },
340 	[TCA_TUNNEL_KEY_ENC_OPTS]     = { .type = NLA_NESTED },
341 	[TCA_TUNNEL_KEY_ENC_TOS]      = { .type = NLA_U8 },
342 	[TCA_TUNNEL_KEY_ENC_TTL]      = { .type = NLA_U8 },
343 };
344 
345 static void tunnel_key_release_params(struct tcf_tunnel_key_params *p)
346 {
347 	if (!p)
348 		return;
349 	if (p->tcft_action == TCA_TUNNEL_KEY_ACT_SET)
350 		dst_release(&p->tcft_enc_metadata->dst);
351 
352 	kfree_rcu(p, rcu);
353 }
354 
355 static int tunnel_key_init(struct net *net, struct nlattr *nla,
356 			   struct nlattr *est, struct tc_action **a,
357 			   int ovr, int bind, bool rtnl_held,
358 			   struct tcf_proto *tp, u32 act_flags,
359 			   struct netlink_ext_ack *extack)
360 {
361 	struct tc_action_net *tn = net_generic(net, tunnel_key_net_id);
362 	struct nlattr *tb[TCA_TUNNEL_KEY_MAX + 1];
363 	struct tcf_tunnel_key_params *params_new;
364 	struct metadata_dst *metadata = NULL;
365 	struct tcf_chain *goto_ch = NULL;
366 	struct tc_tunnel_key *parm;
367 	struct tcf_tunnel_key *t;
368 	bool exists = false;
369 	__be16 dst_port = 0;
370 	__be64 key_id = 0;
371 	int opts_len = 0;
372 	__be16 flags = 0;
373 	u8 tos, ttl;
374 	int ret = 0;
375 	u32 index;
376 	int err;
377 
378 	if (!nla) {
379 		NL_SET_ERR_MSG(extack, "Tunnel requires attributes to be passed");
380 		return -EINVAL;
381 	}
382 
383 	err = nla_parse_nested_deprecated(tb, TCA_TUNNEL_KEY_MAX, nla,
384 					  tunnel_key_policy, extack);
385 	if (err < 0) {
386 		NL_SET_ERR_MSG(extack, "Failed to parse nested tunnel key attributes");
387 		return err;
388 	}
389 
390 	if (!tb[TCA_TUNNEL_KEY_PARMS]) {
391 		NL_SET_ERR_MSG(extack, "Missing tunnel key parameters");
392 		return -EINVAL;
393 	}
394 
395 	parm = nla_data(tb[TCA_TUNNEL_KEY_PARMS]);
396 	index = parm->index;
397 	err = tcf_idr_check_alloc(tn, &index, a, bind);
398 	if (err < 0)
399 		return err;
400 	exists = err;
401 	if (exists && bind)
402 		return 0;
403 
404 	switch (parm->t_action) {
405 	case TCA_TUNNEL_KEY_ACT_RELEASE:
406 		break;
407 	case TCA_TUNNEL_KEY_ACT_SET:
408 		if (tb[TCA_TUNNEL_KEY_ENC_KEY_ID]) {
409 			__be32 key32;
410 
411 			key32 = nla_get_be32(tb[TCA_TUNNEL_KEY_ENC_KEY_ID]);
412 			key_id = key32_to_tunnel_id(key32);
413 			flags = TUNNEL_KEY;
414 		}
415 
416 		flags |= TUNNEL_CSUM;
417 		if (tb[TCA_TUNNEL_KEY_NO_CSUM] &&
418 		    nla_get_u8(tb[TCA_TUNNEL_KEY_NO_CSUM]))
419 			flags &= ~TUNNEL_CSUM;
420 
421 		if (tb[TCA_TUNNEL_KEY_ENC_DST_PORT])
422 			dst_port = nla_get_be16(tb[TCA_TUNNEL_KEY_ENC_DST_PORT]);
423 
424 		if (tb[TCA_TUNNEL_KEY_ENC_OPTS]) {
425 			opts_len = tunnel_key_get_opts_len(tb[TCA_TUNNEL_KEY_ENC_OPTS],
426 							   extack);
427 			if (opts_len < 0) {
428 				ret = opts_len;
429 				goto err_out;
430 			}
431 		}
432 
433 		tos = 0;
434 		if (tb[TCA_TUNNEL_KEY_ENC_TOS])
435 			tos = nla_get_u8(tb[TCA_TUNNEL_KEY_ENC_TOS]);
436 		ttl = 0;
437 		if (tb[TCA_TUNNEL_KEY_ENC_TTL])
438 			ttl = nla_get_u8(tb[TCA_TUNNEL_KEY_ENC_TTL]);
439 
440 		if (tb[TCA_TUNNEL_KEY_ENC_IPV4_SRC] &&
441 		    tb[TCA_TUNNEL_KEY_ENC_IPV4_DST]) {
442 			__be32 saddr;
443 			__be32 daddr;
444 
445 			saddr = nla_get_in_addr(tb[TCA_TUNNEL_KEY_ENC_IPV4_SRC]);
446 			daddr = nla_get_in_addr(tb[TCA_TUNNEL_KEY_ENC_IPV4_DST]);
447 
448 			metadata = __ip_tun_set_dst(saddr, daddr, tos, ttl,
449 						    dst_port, flags,
450 						    key_id, opts_len);
451 		} else if (tb[TCA_TUNNEL_KEY_ENC_IPV6_SRC] &&
452 			   tb[TCA_TUNNEL_KEY_ENC_IPV6_DST]) {
453 			struct in6_addr saddr;
454 			struct in6_addr daddr;
455 
456 			saddr = nla_get_in6_addr(tb[TCA_TUNNEL_KEY_ENC_IPV6_SRC]);
457 			daddr = nla_get_in6_addr(tb[TCA_TUNNEL_KEY_ENC_IPV6_DST]);
458 
459 			metadata = __ipv6_tun_set_dst(&saddr, &daddr, tos, ttl, dst_port,
460 						      0, flags,
461 						      key_id, 0);
462 		} else {
463 			NL_SET_ERR_MSG(extack, "Missing either ipv4 or ipv6 src and dst");
464 			ret = -EINVAL;
465 			goto err_out;
466 		}
467 
468 		if (!metadata) {
469 			NL_SET_ERR_MSG(extack, "Cannot allocate tunnel metadata dst");
470 			ret = -ENOMEM;
471 			goto err_out;
472 		}
473 
474 #ifdef CONFIG_DST_CACHE
475 		ret = dst_cache_init(&metadata->u.tun_info.dst_cache, GFP_KERNEL);
476 		if (ret)
477 			goto release_tun_meta;
478 #endif
479 
480 		if (opts_len) {
481 			ret = tunnel_key_opts_set(tb[TCA_TUNNEL_KEY_ENC_OPTS],
482 						  &metadata->u.tun_info,
483 						  opts_len, extack);
484 			if (ret < 0)
485 				goto release_tun_meta;
486 		}
487 
488 		metadata->u.tun_info.mode |= IP_TUNNEL_INFO_TX;
489 		break;
490 	default:
491 		NL_SET_ERR_MSG(extack, "Unknown tunnel key action");
492 		ret = -EINVAL;
493 		goto err_out;
494 	}
495 
496 	if (!exists) {
497 		ret = tcf_idr_create_from_flags(tn, index, est, a,
498 						&act_tunnel_key_ops, bind,
499 						act_flags);
500 		if (ret) {
501 			NL_SET_ERR_MSG(extack, "Cannot create TC IDR");
502 			goto release_tun_meta;
503 		}
504 
505 		ret = ACT_P_CREATED;
506 	} else if (!ovr) {
507 		NL_SET_ERR_MSG(extack, "TC IDR already exists");
508 		ret = -EEXIST;
509 		goto release_tun_meta;
510 	}
511 
512 	err = tcf_action_check_ctrlact(parm->action, tp, &goto_ch, extack);
513 	if (err < 0) {
514 		ret = err;
515 		exists = true;
516 		goto release_tun_meta;
517 	}
518 	t = to_tunnel_key(*a);
519 
520 	params_new = kzalloc(sizeof(*params_new), GFP_KERNEL);
521 	if (unlikely(!params_new)) {
522 		NL_SET_ERR_MSG(extack, "Cannot allocate tunnel key parameters");
523 		ret = -ENOMEM;
524 		exists = true;
525 		goto put_chain;
526 	}
527 	params_new->tcft_action = parm->t_action;
528 	params_new->tcft_enc_metadata = metadata;
529 
530 	spin_lock_bh(&t->tcf_lock);
531 	goto_ch = tcf_action_set_ctrlact(*a, parm->action, goto_ch);
532 	params_new = rcu_replace_pointer(t->params, params_new,
533 					 lockdep_is_held(&t->tcf_lock));
534 	spin_unlock_bh(&t->tcf_lock);
535 	tunnel_key_release_params(params_new);
536 	if (goto_ch)
537 		tcf_chain_put_by_act(goto_ch);
538 
539 	if (ret == ACT_P_CREATED)
540 		tcf_idr_insert(tn, *a);
541 
542 	return ret;
543 
544 put_chain:
545 	if (goto_ch)
546 		tcf_chain_put_by_act(goto_ch);
547 
548 release_tun_meta:
549 	if (metadata)
550 		dst_release(&metadata->dst);
551 
552 err_out:
553 	if (exists)
554 		tcf_idr_release(*a, bind);
555 	else
556 		tcf_idr_cleanup(tn, index);
557 	return ret;
558 }
559 
560 static void tunnel_key_release(struct tc_action *a)
561 {
562 	struct tcf_tunnel_key *t = to_tunnel_key(a);
563 	struct tcf_tunnel_key_params *params;
564 
565 	params = rcu_dereference_protected(t->params, 1);
566 	tunnel_key_release_params(params);
567 }
568 
569 static int tunnel_key_geneve_opts_dump(struct sk_buff *skb,
570 				       const struct ip_tunnel_info *info)
571 {
572 	int len = info->options_len;
573 	u8 *src = (u8 *)(info + 1);
574 	struct nlattr *start;
575 
576 	start = nla_nest_start_noflag(skb, TCA_TUNNEL_KEY_ENC_OPTS_GENEVE);
577 	if (!start)
578 		return -EMSGSIZE;
579 
580 	while (len > 0) {
581 		struct geneve_opt *opt = (struct geneve_opt *)src;
582 
583 		if (nla_put_be16(skb, TCA_TUNNEL_KEY_ENC_OPT_GENEVE_CLASS,
584 				 opt->opt_class) ||
585 		    nla_put_u8(skb, TCA_TUNNEL_KEY_ENC_OPT_GENEVE_TYPE,
586 			       opt->type) ||
587 		    nla_put(skb, TCA_TUNNEL_KEY_ENC_OPT_GENEVE_DATA,
588 			    opt->length * 4, opt + 1)) {
589 			nla_nest_cancel(skb, start);
590 			return -EMSGSIZE;
591 		}
592 
593 		len -= sizeof(struct geneve_opt) + opt->length * 4;
594 		src += sizeof(struct geneve_opt) + opt->length * 4;
595 	}
596 
597 	nla_nest_end(skb, start);
598 	return 0;
599 }
600 
601 static int tunnel_key_vxlan_opts_dump(struct sk_buff *skb,
602 				      const struct ip_tunnel_info *info)
603 {
604 	struct vxlan_metadata *md = (struct vxlan_metadata *)(info + 1);
605 	struct nlattr *start;
606 
607 	start = nla_nest_start_noflag(skb, TCA_TUNNEL_KEY_ENC_OPTS_VXLAN);
608 	if (!start)
609 		return -EMSGSIZE;
610 
611 	if (nla_put_u32(skb, TCA_TUNNEL_KEY_ENC_OPT_VXLAN_GBP, md->gbp)) {
612 		nla_nest_cancel(skb, start);
613 		return -EMSGSIZE;
614 	}
615 
616 	nla_nest_end(skb, start);
617 	return 0;
618 }
619 
620 static int tunnel_key_erspan_opts_dump(struct sk_buff *skb,
621 				       const struct ip_tunnel_info *info)
622 {
623 	struct erspan_metadata *md = (struct erspan_metadata *)(info + 1);
624 	struct nlattr *start;
625 
626 	start = nla_nest_start_noflag(skb, TCA_TUNNEL_KEY_ENC_OPTS_ERSPAN);
627 	if (!start)
628 		return -EMSGSIZE;
629 
630 	if (nla_put_u8(skb, TCA_TUNNEL_KEY_ENC_OPT_ERSPAN_VER, md->version))
631 		goto err;
632 
633 	if (md->version == 1 &&
634 	    nla_put_be32(skb, TCA_TUNNEL_KEY_ENC_OPT_ERSPAN_INDEX, md->u.index))
635 		goto err;
636 
637 	if (md->version == 2 &&
638 	    (nla_put_u8(skb, TCA_TUNNEL_KEY_ENC_OPT_ERSPAN_DIR,
639 			md->u.md2.dir) ||
640 	     nla_put_u8(skb, TCA_TUNNEL_KEY_ENC_OPT_ERSPAN_HWID,
641 			get_hwid(&md->u.md2))))
642 		goto err;
643 
644 	nla_nest_end(skb, start);
645 	return 0;
646 err:
647 	nla_nest_cancel(skb, start);
648 	return -EMSGSIZE;
649 }
650 
651 static int tunnel_key_opts_dump(struct sk_buff *skb,
652 				const struct ip_tunnel_info *info)
653 {
654 	struct nlattr *start;
655 	int err = -EINVAL;
656 
657 	if (!info->options_len)
658 		return 0;
659 
660 	start = nla_nest_start_noflag(skb, TCA_TUNNEL_KEY_ENC_OPTS);
661 	if (!start)
662 		return -EMSGSIZE;
663 
664 	if (info->key.tun_flags & TUNNEL_GENEVE_OPT) {
665 		err = tunnel_key_geneve_opts_dump(skb, info);
666 		if (err)
667 			goto err_out;
668 	} else if (info->key.tun_flags & TUNNEL_VXLAN_OPT) {
669 		err = tunnel_key_vxlan_opts_dump(skb, info);
670 		if (err)
671 			goto err_out;
672 	} else if (info->key.tun_flags & TUNNEL_ERSPAN_OPT) {
673 		err = tunnel_key_erspan_opts_dump(skb, info);
674 		if (err)
675 			goto err_out;
676 	} else {
677 err_out:
678 		nla_nest_cancel(skb, start);
679 		return err;
680 	}
681 
682 	nla_nest_end(skb, start);
683 	return 0;
684 }
685 
686 static int tunnel_key_dump_addresses(struct sk_buff *skb,
687 				     const struct ip_tunnel_info *info)
688 {
689 	unsigned short family = ip_tunnel_info_af(info);
690 
691 	if (family == AF_INET) {
692 		__be32 saddr = info->key.u.ipv4.src;
693 		__be32 daddr = info->key.u.ipv4.dst;
694 
695 		if (!nla_put_in_addr(skb, TCA_TUNNEL_KEY_ENC_IPV4_SRC, saddr) &&
696 		    !nla_put_in_addr(skb, TCA_TUNNEL_KEY_ENC_IPV4_DST, daddr))
697 			return 0;
698 	}
699 
700 	if (family == AF_INET6) {
701 		const struct in6_addr *saddr6 = &info->key.u.ipv6.src;
702 		const struct in6_addr *daddr6 = &info->key.u.ipv6.dst;
703 
704 		if (!nla_put_in6_addr(skb,
705 				      TCA_TUNNEL_KEY_ENC_IPV6_SRC, saddr6) &&
706 		    !nla_put_in6_addr(skb,
707 				      TCA_TUNNEL_KEY_ENC_IPV6_DST, daddr6))
708 			return 0;
709 	}
710 
711 	return -EINVAL;
712 }
713 
714 static int tunnel_key_dump(struct sk_buff *skb, struct tc_action *a,
715 			   int bind, int ref)
716 {
717 	unsigned char *b = skb_tail_pointer(skb);
718 	struct tcf_tunnel_key *t = to_tunnel_key(a);
719 	struct tcf_tunnel_key_params *params;
720 	struct tc_tunnel_key opt = {
721 		.index    = t->tcf_index,
722 		.refcnt   = refcount_read(&t->tcf_refcnt) - ref,
723 		.bindcnt  = atomic_read(&t->tcf_bindcnt) - bind,
724 	};
725 	struct tcf_t tm;
726 
727 	spin_lock_bh(&t->tcf_lock);
728 	params = rcu_dereference_protected(t->params,
729 					   lockdep_is_held(&t->tcf_lock));
730 	opt.action   = t->tcf_action;
731 	opt.t_action = params->tcft_action;
732 
733 	if (nla_put(skb, TCA_TUNNEL_KEY_PARMS, sizeof(opt), &opt))
734 		goto nla_put_failure;
735 
736 	if (params->tcft_action == TCA_TUNNEL_KEY_ACT_SET) {
737 		struct ip_tunnel_info *info =
738 			&params->tcft_enc_metadata->u.tun_info;
739 		struct ip_tunnel_key *key = &info->key;
740 		__be32 key_id = tunnel_id_to_key32(key->tun_id);
741 
742 		if (((key->tun_flags & TUNNEL_KEY) &&
743 		     nla_put_be32(skb, TCA_TUNNEL_KEY_ENC_KEY_ID, key_id)) ||
744 		    tunnel_key_dump_addresses(skb,
745 					      &params->tcft_enc_metadata->u.tun_info) ||
746 		    (key->tp_dst &&
747 		      nla_put_be16(skb, TCA_TUNNEL_KEY_ENC_DST_PORT,
748 				   key->tp_dst)) ||
749 		    nla_put_u8(skb, TCA_TUNNEL_KEY_NO_CSUM,
750 			       !(key->tun_flags & TUNNEL_CSUM)) ||
751 		    tunnel_key_opts_dump(skb, info))
752 			goto nla_put_failure;
753 
754 		if (key->tos && nla_put_u8(skb, TCA_TUNNEL_KEY_ENC_TOS, key->tos))
755 			goto nla_put_failure;
756 
757 		if (key->ttl && nla_put_u8(skb, TCA_TUNNEL_KEY_ENC_TTL, key->ttl))
758 			goto nla_put_failure;
759 	}
760 
761 	tcf_tm_dump(&tm, &t->tcf_tm);
762 	if (nla_put_64bit(skb, TCA_TUNNEL_KEY_TM, sizeof(tm),
763 			  &tm, TCA_TUNNEL_KEY_PAD))
764 		goto nla_put_failure;
765 	spin_unlock_bh(&t->tcf_lock);
766 
767 	return skb->len;
768 
769 nla_put_failure:
770 	spin_unlock_bh(&t->tcf_lock);
771 	nlmsg_trim(skb, b);
772 	return -1;
773 }
774 
775 static int tunnel_key_walker(struct net *net, struct sk_buff *skb,
776 			     struct netlink_callback *cb, int type,
777 			     const struct tc_action_ops *ops,
778 			     struct netlink_ext_ack *extack)
779 {
780 	struct tc_action_net *tn = net_generic(net, tunnel_key_net_id);
781 
782 	return tcf_generic_walker(tn, skb, cb, type, ops, extack);
783 }
784 
785 static int tunnel_key_search(struct net *net, struct tc_action **a, u32 index)
786 {
787 	struct tc_action_net *tn = net_generic(net, tunnel_key_net_id);
788 
789 	return tcf_idr_search(tn, a, index);
790 }
791 
792 static struct tc_action_ops act_tunnel_key_ops = {
793 	.kind		=	"tunnel_key",
794 	.id		=	TCA_ID_TUNNEL_KEY,
795 	.owner		=	THIS_MODULE,
796 	.act		=	tunnel_key_act,
797 	.dump		=	tunnel_key_dump,
798 	.init		=	tunnel_key_init,
799 	.cleanup	=	tunnel_key_release,
800 	.walk		=	tunnel_key_walker,
801 	.lookup		=	tunnel_key_search,
802 	.size		=	sizeof(struct tcf_tunnel_key),
803 };
804 
805 static __net_init int tunnel_key_init_net(struct net *net)
806 {
807 	struct tc_action_net *tn = net_generic(net, tunnel_key_net_id);
808 
809 	return tc_action_net_init(net, tn, &act_tunnel_key_ops);
810 }
811 
812 static void __net_exit tunnel_key_exit_net(struct list_head *net_list)
813 {
814 	tc_action_net_exit(net_list, tunnel_key_net_id);
815 }
816 
817 static struct pernet_operations tunnel_key_net_ops = {
818 	.init = tunnel_key_init_net,
819 	.exit_batch = tunnel_key_exit_net,
820 	.id   = &tunnel_key_net_id,
821 	.size = sizeof(struct tc_action_net),
822 };
823 
824 static int __init tunnel_key_init_module(void)
825 {
826 	return tcf_register_action(&act_tunnel_key_ops, &tunnel_key_net_ops);
827 }
828 
829 static void __exit tunnel_key_cleanup_module(void)
830 {
831 	tcf_unregister_action(&act_tunnel_key_ops, &tunnel_key_net_ops);
832 }
833 
834 module_init(tunnel_key_init_module);
835 module_exit(tunnel_key_cleanup_module);
836 
837 MODULE_AUTHOR("Amir Vadai <amir@vadai.me>");
838 MODULE_DESCRIPTION("ip tunnel manipulation actions");
839 MODULE_LICENSE("GPL v2");
840