xref: /openbmc/linux/net/sched/act_tunnel_key.c (revision ae108c48)
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 struct tc_action_ops act_tunnel_key_ops;
24 
25 static int tunnel_key_act(struct sk_buff *skb, const struct tc_action *a,
26 			  struct tcf_result *res)
27 {
28 	struct tcf_tunnel_key *t = to_tunnel_key(a);
29 	struct tcf_tunnel_key_params *params;
30 	int action;
31 
32 	params = rcu_dereference_bh(t->params);
33 
34 	tcf_lastuse_update(&t->tcf_tm);
35 	tcf_action_update_bstats(&t->common, skb);
36 	action = READ_ONCE(t->tcf_action);
37 
38 	switch (params->tcft_action) {
39 	case TCA_TUNNEL_KEY_ACT_RELEASE:
40 		skb_dst_drop(skb);
41 		break;
42 	case TCA_TUNNEL_KEY_ACT_SET:
43 		skb_dst_drop(skb);
44 		skb_dst_set(skb, dst_clone(&params->tcft_enc_metadata->dst));
45 		break;
46 	default:
47 		WARN_ONCE(1, "Bad tunnel_key action %d.\n",
48 			  params->tcft_action);
49 		break;
50 	}
51 
52 	return action;
53 }
54 
55 static const struct nla_policy
56 enc_opts_policy[TCA_TUNNEL_KEY_ENC_OPTS_MAX + 1] = {
57 	[TCA_TUNNEL_KEY_ENC_OPTS_UNSPEC]	= {
58 		.strict_start_type = TCA_TUNNEL_KEY_ENC_OPTS_VXLAN },
59 	[TCA_TUNNEL_KEY_ENC_OPTS_GENEVE]	= { .type = NLA_NESTED },
60 	[TCA_TUNNEL_KEY_ENC_OPTS_VXLAN]		= { .type = NLA_NESTED },
61 	[TCA_TUNNEL_KEY_ENC_OPTS_ERSPAN]	= { .type = NLA_NESTED },
62 };
63 
64 static const struct nla_policy
65 geneve_opt_policy[TCA_TUNNEL_KEY_ENC_OPT_GENEVE_MAX + 1] = {
66 	[TCA_TUNNEL_KEY_ENC_OPT_GENEVE_CLASS]	   = { .type = NLA_U16 },
67 	[TCA_TUNNEL_KEY_ENC_OPT_GENEVE_TYPE]	   = { .type = NLA_U8 },
68 	[TCA_TUNNEL_KEY_ENC_OPT_GENEVE_DATA]	   = { .type = NLA_BINARY,
69 						       .len = 128 },
70 };
71 
72 static const struct nla_policy
73 vxlan_opt_policy[TCA_TUNNEL_KEY_ENC_OPT_VXLAN_MAX + 1] = {
74 	[TCA_TUNNEL_KEY_ENC_OPT_VXLAN_GBP]	   = { .type = NLA_U32 },
75 };
76 
77 static const struct nla_policy
78 erspan_opt_policy[TCA_TUNNEL_KEY_ENC_OPT_ERSPAN_MAX + 1] = {
79 	[TCA_TUNNEL_KEY_ENC_OPT_ERSPAN_VER]	   = { .type = NLA_U8 },
80 	[TCA_TUNNEL_KEY_ENC_OPT_ERSPAN_INDEX]	   = { .type = NLA_U32 },
81 	[TCA_TUNNEL_KEY_ENC_OPT_ERSPAN_DIR]	   = { .type = NLA_U8 },
82 	[TCA_TUNNEL_KEY_ENC_OPT_ERSPAN_HWID]	   = { .type = NLA_U8 },
83 };
84 
85 static int
86 tunnel_key_copy_geneve_opt(const struct nlattr *nla, void *dst, int dst_len,
87 			   struct netlink_ext_ack *extack)
88 {
89 	struct nlattr *tb[TCA_TUNNEL_KEY_ENC_OPT_GENEVE_MAX + 1];
90 	int err, data_len, opt_len;
91 	u8 *data;
92 
93 	err = nla_parse_nested_deprecated(tb,
94 					  TCA_TUNNEL_KEY_ENC_OPT_GENEVE_MAX,
95 					  nla, geneve_opt_policy, extack);
96 	if (err < 0)
97 		return err;
98 
99 	if (!tb[TCA_TUNNEL_KEY_ENC_OPT_GENEVE_CLASS] ||
100 	    !tb[TCA_TUNNEL_KEY_ENC_OPT_GENEVE_TYPE] ||
101 	    !tb[TCA_TUNNEL_KEY_ENC_OPT_GENEVE_DATA]) {
102 		NL_SET_ERR_MSG(extack, "Missing tunnel key geneve option class, type or data");
103 		return -EINVAL;
104 	}
105 
106 	data = nla_data(tb[TCA_TUNNEL_KEY_ENC_OPT_GENEVE_DATA]);
107 	data_len = nla_len(tb[TCA_TUNNEL_KEY_ENC_OPT_GENEVE_DATA]);
108 	if (data_len < 4) {
109 		NL_SET_ERR_MSG(extack, "Tunnel key geneve option data is less than 4 bytes long");
110 		return -ERANGE;
111 	}
112 	if (data_len % 4) {
113 		NL_SET_ERR_MSG(extack, "Tunnel key geneve option data is not a multiple of 4 bytes long");
114 		return -ERANGE;
115 	}
116 
117 	opt_len = sizeof(struct geneve_opt) + data_len;
118 	if (dst) {
119 		struct geneve_opt *opt = dst;
120 
121 		WARN_ON(dst_len < opt_len);
122 
123 		opt->opt_class =
124 			nla_get_be16(tb[TCA_TUNNEL_KEY_ENC_OPT_GENEVE_CLASS]);
125 		opt->type = nla_get_u8(tb[TCA_TUNNEL_KEY_ENC_OPT_GENEVE_TYPE]);
126 		opt->length = data_len / 4; /* length is in units of 4 bytes */
127 		opt->r1 = 0;
128 		opt->r2 = 0;
129 		opt->r3 = 0;
130 
131 		memcpy(opt + 1, data, data_len);
132 	}
133 
134 	return opt_len;
135 }
136 
137 static int
138 tunnel_key_copy_vxlan_opt(const struct nlattr *nla, void *dst, int dst_len,
139 			  struct netlink_ext_ack *extack)
140 {
141 	struct nlattr *tb[TCA_TUNNEL_KEY_ENC_OPT_VXLAN_MAX + 1];
142 	int err;
143 
144 	err = nla_parse_nested(tb, TCA_TUNNEL_KEY_ENC_OPT_VXLAN_MAX, nla,
145 			       vxlan_opt_policy, extack);
146 	if (err < 0)
147 		return err;
148 
149 	if (!tb[TCA_TUNNEL_KEY_ENC_OPT_VXLAN_GBP]) {
150 		NL_SET_ERR_MSG(extack, "Missing tunnel key vxlan option gbp");
151 		return -EINVAL;
152 	}
153 
154 	if (dst) {
155 		struct vxlan_metadata *md = dst;
156 
157 		md->gbp = nla_get_u32(tb[TCA_TUNNEL_KEY_ENC_OPT_VXLAN_GBP]);
158 		md->gbp &= VXLAN_GBP_MASK;
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 			   struct tcf_proto *tp, u32 act_flags,
358 			   struct netlink_ext_ack *extack)
359 {
360 	struct tc_action_net *tn = net_generic(net, act_tunnel_key_ops.net_id);
361 	bool bind = act_flags & TCA_ACT_FLAGS_BIND;
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, opts_len);
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 (!(act_flags & TCA_ACT_FLAGS_REPLACE)) {
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 	return ret;
540 
541 put_chain:
542 	if (goto_ch)
543 		tcf_chain_put_by_act(goto_ch);
544 
545 release_tun_meta:
546 	if (metadata)
547 		dst_release(&metadata->dst);
548 
549 err_out:
550 	if (exists)
551 		tcf_idr_release(*a, bind);
552 	else
553 		tcf_idr_cleanup(tn, index);
554 	return ret;
555 }
556 
557 static void tunnel_key_release(struct tc_action *a)
558 {
559 	struct tcf_tunnel_key *t = to_tunnel_key(a);
560 	struct tcf_tunnel_key_params *params;
561 
562 	params = rcu_dereference_protected(t->params, 1);
563 	tunnel_key_release_params(params);
564 }
565 
566 static int tunnel_key_geneve_opts_dump(struct sk_buff *skb,
567 				       const struct ip_tunnel_info *info)
568 {
569 	int len = info->options_len;
570 	u8 *src = (u8 *)(info + 1);
571 	struct nlattr *start;
572 
573 	start = nla_nest_start_noflag(skb, TCA_TUNNEL_KEY_ENC_OPTS_GENEVE);
574 	if (!start)
575 		return -EMSGSIZE;
576 
577 	while (len > 0) {
578 		struct geneve_opt *opt = (struct geneve_opt *)src;
579 
580 		if (nla_put_be16(skb, TCA_TUNNEL_KEY_ENC_OPT_GENEVE_CLASS,
581 				 opt->opt_class) ||
582 		    nla_put_u8(skb, TCA_TUNNEL_KEY_ENC_OPT_GENEVE_TYPE,
583 			       opt->type) ||
584 		    nla_put(skb, TCA_TUNNEL_KEY_ENC_OPT_GENEVE_DATA,
585 			    opt->length * 4, opt + 1)) {
586 			nla_nest_cancel(skb, start);
587 			return -EMSGSIZE;
588 		}
589 
590 		len -= sizeof(struct geneve_opt) + opt->length * 4;
591 		src += sizeof(struct geneve_opt) + opt->length * 4;
592 	}
593 
594 	nla_nest_end(skb, start);
595 	return 0;
596 }
597 
598 static int tunnel_key_vxlan_opts_dump(struct sk_buff *skb,
599 				      const struct ip_tunnel_info *info)
600 {
601 	struct vxlan_metadata *md = (struct vxlan_metadata *)(info + 1);
602 	struct nlattr *start;
603 
604 	start = nla_nest_start_noflag(skb, TCA_TUNNEL_KEY_ENC_OPTS_VXLAN);
605 	if (!start)
606 		return -EMSGSIZE;
607 
608 	if (nla_put_u32(skb, TCA_TUNNEL_KEY_ENC_OPT_VXLAN_GBP, md->gbp)) {
609 		nla_nest_cancel(skb, start);
610 		return -EMSGSIZE;
611 	}
612 
613 	nla_nest_end(skb, start);
614 	return 0;
615 }
616 
617 static int tunnel_key_erspan_opts_dump(struct sk_buff *skb,
618 				       const struct ip_tunnel_info *info)
619 {
620 	struct erspan_metadata *md = (struct erspan_metadata *)(info + 1);
621 	struct nlattr *start;
622 
623 	start = nla_nest_start_noflag(skb, TCA_TUNNEL_KEY_ENC_OPTS_ERSPAN);
624 	if (!start)
625 		return -EMSGSIZE;
626 
627 	if (nla_put_u8(skb, TCA_TUNNEL_KEY_ENC_OPT_ERSPAN_VER, md->version))
628 		goto err;
629 
630 	if (md->version == 1 &&
631 	    nla_put_be32(skb, TCA_TUNNEL_KEY_ENC_OPT_ERSPAN_INDEX, md->u.index))
632 		goto err;
633 
634 	if (md->version == 2 &&
635 	    (nla_put_u8(skb, TCA_TUNNEL_KEY_ENC_OPT_ERSPAN_DIR,
636 			md->u.md2.dir) ||
637 	     nla_put_u8(skb, TCA_TUNNEL_KEY_ENC_OPT_ERSPAN_HWID,
638 			get_hwid(&md->u.md2))))
639 		goto err;
640 
641 	nla_nest_end(skb, start);
642 	return 0;
643 err:
644 	nla_nest_cancel(skb, start);
645 	return -EMSGSIZE;
646 }
647 
648 static int tunnel_key_opts_dump(struct sk_buff *skb,
649 				const struct ip_tunnel_info *info)
650 {
651 	struct nlattr *start;
652 	int err = -EINVAL;
653 
654 	if (!info->options_len)
655 		return 0;
656 
657 	start = nla_nest_start_noflag(skb, TCA_TUNNEL_KEY_ENC_OPTS);
658 	if (!start)
659 		return -EMSGSIZE;
660 
661 	if (info->key.tun_flags & TUNNEL_GENEVE_OPT) {
662 		err = tunnel_key_geneve_opts_dump(skb, info);
663 		if (err)
664 			goto err_out;
665 	} else if (info->key.tun_flags & TUNNEL_VXLAN_OPT) {
666 		err = tunnel_key_vxlan_opts_dump(skb, info);
667 		if (err)
668 			goto err_out;
669 	} else if (info->key.tun_flags & TUNNEL_ERSPAN_OPT) {
670 		err = tunnel_key_erspan_opts_dump(skb, info);
671 		if (err)
672 			goto err_out;
673 	} else {
674 err_out:
675 		nla_nest_cancel(skb, start);
676 		return err;
677 	}
678 
679 	nla_nest_end(skb, start);
680 	return 0;
681 }
682 
683 static int tunnel_key_dump_addresses(struct sk_buff *skb,
684 				     const struct ip_tunnel_info *info)
685 {
686 	unsigned short family = ip_tunnel_info_af(info);
687 
688 	if (family == AF_INET) {
689 		__be32 saddr = info->key.u.ipv4.src;
690 		__be32 daddr = info->key.u.ipv4.dst;
691 
692 		if (!nla_put_in_addr(skb, TCA_TUNNEL_KEY_ENC_IPV4_SRC, saddr) &&
693 		    !nla_put_in_addr(skb, TCA_TUNNEL_KEY_ENC_IPV4_DST, daddr))
694 			return 0;
695 	}
696 
697 	if (family == AF_INET6) {
698 		const struct in6_addr *saddr6 = &info->key.u.ipv6.src;
699 		const struct in6_addr *daddr6 = &info->key.u.ipv6.dst;
700 
701 		if (!nla_put_in6_addr(skb,
702 				      TCA_TUNNEL_KEY_ENC_IPV6_SRC, saddr6) &&
703 		    !nla_put_in6_addr(skb,
704 				      TCA_TUNNEL_KEY_ENC_IPV6_DST, daddr6))
705 			return 0;
706 	}
707 
708 	return -EINVAL;
709 }
710 
711 static int tunnel_key_dump(struct sk_buff *skb, struct tc_action *a,
712 			   int bind, int ref)
713 {
714 	unsigned char *b = skb_tail_pointer(skb);
715 	struct tcf_tunnel_key *t = to_tunnel_key(a);
716 	struct tcf_tunnel_key_params *params;
717 	struct tc_tunnel_key opt = {
718 		.index    = t->tcf_index,
719 		.refcnt   = refcount_read(&t->tcf_refcnt) - ref,
720 		.bindcnt  = atomic_read(&t->tcf_bindcnt) - bind,
721 	};
722 	struct tcf_t tm;
723 
724 	spin_lock_bh(&t->tcf_lock);
725 	params = rcu_dereference_protected(t->params,
726 					   lockdep_is_held(&t->tcf_lock));
727 	opt.action   = t->tcf_action;
728 	opt.t_action = params->tcft_action;
729 
730 	if (nla_put(skb, TCA_TUNNEL_KEY_PARMS, sizeof(opt), &opt))
731 		goto nla_put_failure;
732 
733 	if (params->tcft_action == TCA_TUNNEL_KEY_ACT_SET) {
734 		struct ip_tunnel_info *info =
735 			&params->tcft_enc_metadata->u.tun_info;
736 		struct ip_tunnel_key *key = &info->key;
737 		__be32 key_id = tunnel_id_to_key32(key->tun_id);
738 
739 		if (((key->tun_flags & TUNNEL_KEY) &&
740 		     nla_put_be32(skb, TCA_TUNNEL_KEY_ENC_KEY_ID, key_id)) ||
741 		    tunnel_key_dump_addresses(skb,
742 					      &params->tcft_enc_metadata->u.tun_info) ||
743 		    (key->tp_dst &&
744 		      nla_put_be16(skb, TCA_TUNNEL_KEY_ENC_DST_PORT,
745 				   key->tp_dst)) ||
746 		    nla_put_u8(skb, TCA_TUNNEL_KEY_NO_CSUM,
747 			       !(key->tun_flags & TUNNEL_CSUM)) ||
748 		    tunnel_key_opts_dump(skb, info))
749 			goto nla_put_failure;
750 
751 		if (key->tos && nla_put_u8(skb, TCA_TUNNEL_KEY_ENC_TOS, key->tos))
752 			goto nla_put_failure;
753 
754 		if (key->ttl && nla_put_u8(skb, TCA_TUNNEL_KEY_ENC_TTL, key->ttl))
755 			goto nla_put_failure;
756 	}
757 
758 	tcf_tm_dump(&tm, &t->tcf_tm);
759 	if (nla_put_64bit(skb, TCA_TUNNEL_KEY_TM, sizeof(tm),
760 			  &tm, TCA_TUNNEL_KEY_PAD))
761 		goto nla_put_failure;
762 	spin_unlock_bh(&t->tcf_lock);
763 
764 	return skb->len;
765 
766 nla_put_failure:
767 	spin_unlock_bh(&t->tcf_lock);
768 	nlmsg_trim(skb, b);
769 	return -1;
770 }
771 
772 static void tcf_tunnel_encap_put_tunnel(void *priv)
773 {
774 	struct ip_tunnel_info *tunnel = priv;
775 
776 	kfree(tunnel);
777 }
778 
779 static int tcf_tunnel_encap_get_tunnel(struct flow_action_entry *entry,
780 				       const struct tc_action *act)
781 {
782 	entry->tunnel = tcf_tunnel_info_copy(act);
783 	if (!entry->tunnel)
784 		return -ENOMEM;
785 	entry->destructor = tcf_tunnel_encap_put_tunnel;
786 	entry->destructor_priv = entry->tunnel;
787 	return 0;
788 }
789 
790 static int tcf_tunnel_key_offload_act_setup(struct tc_action *act,
791 					    void *entry_data,
792 					    u32 *index_inc,
793 					    bool bind,
794 					    struct netlink_ext_ack *extack)
795 {
796 	int err;
797 
798 	if (bind) {
799 		struct flow_action_entry *entry = entry_data;
800 
801 		if (is_tcf_tunnel_set(act)) {
802 			entry->id = FLOW_ACTION_TUNNEL_ENCAP;
803 			err = tcf_tunnel_encap_get_tunnel(entry, act);
804 			if (err)
805 				return err;
806 		} else if (is_tcf_tunnel_release(act)) {
807 			entry->id = FLOW_ACTION_TUNNEL_DECAP;
808 		} else {
809 			NL_SET_ERR_MSG_MOD(extack, "Unsupported tunnel key mode offload");
810 			return -EOPNOTSUPP;
811 		}
812 		*index_inc = 1;
813 	} else {
814 		struct flow_offload_action *fl_action = entry_data;
815 
816 		if (is_tcf_tunnel_set(act))
817 			fl_action->id = FLOW_ACTION_TUNNEL_ENCAP;
818 		else if (is_tcf_tunnel_release(act))
819 			fl_action->id = FLOW_ACTION_TUNNEL_DECAP;
820 		else
821 			return -EOPNOTSUPP;
822 	}
823 
824 	return 0;
825 }
826 
827 static struct tc_action_ops act_tunnel_key_ops = {
828 	.kind		=	"tunnel_key",
829 	.id		=	TCA_ID_TUNNEL_KEY,
830 	.owner		=	THIS_MODULE,
831 	.act		=	tunnel_key_act,
832 	.dump		=	tunnel_key_dump,
833 	.init		=	tunnel_key_init,
834 	.cleanup	=	tunnel_key_release,
835 	.offload_act_setup =	tcf_tunnel_key_offload_act_setup,
836 	.size		=	sizeof(struct tcf_tunnel_key),
837 };
838 
839 static __net_init int tunnel_key_init_net(struct net *net)
840 {
841 	struct tc_action_net *tn = net_generic(net, act_tunnel_key_ops.net_id);
842 
843 	return tc_action_net_init(net, tn, &act_tunnel_key_ops);
844 }
845 
846 static void __net_exit tunnel_key_exit_net(struct list_head *net_list)
847 {
848 	tc_action_net_exit(net_list, act_tunnel_key_ops.net_id);
849 }
850 
851 static struct pernet_operations tunnel_key_net_ops = {
852 	.init = tunnel_key_init_net,
853 	.exit_batch = tunnel_key_exit_net,
854 	.id   = &act_tunnel_key_ops.net_id,
855 	.size = sizeof(struct tc_action_net),
856 };
857 
858 static int __init tunnel_key_init_module(void)
859 {
860 	return tcf_register_action(&act_tunnel_key_ops, &tunnel_key_net_ops);
861 }
862 
863 static void __exit tunnel_key_cleanup_module(void)
864 {
865 	tcf_unregister_action(&act_tunnel_key_ops, &tunnel_key_net_ops);
866 }
867 
868 module_init(tunnel_key_init_module);
869 module_exit(tunnel_key_cleanup_module);
870 
871 MODULE_AUTHOR("Amir Vadai <amir@vadai.me>");
872 MODULE_DESCRIPTION("ip tunnel manipulation actions");
873 MODULE_LICENSE("GPL v2");
874