xref: /openbmc/linux/net/sched/act_tunnel_key.c (revision 6197e5b7)
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 		md->gbp &= VXLAN_GBP_MASK;
160 	}
161 
162 	return sizeof(struct vxlan_metadata);
163 }
164 
165 static int
166 tunnel_key_copy_erspan_opt(const struct nlattr *nla, void *dst, int dst_len,
167 			   struct netlink_ext_ack *extack)
168 {
169 	struct nlattr *tb[TCA_TUNNEL_KEY_ENC_OPT_ERSPAN_MAX + 1];
170 	int err;
171 	u8 ver;
172 
173 	err = nla_parse_nested(tb, TCA_TUNNEL_KEY_ENC_OPT_ERSPAN_MAX, nla,
174 			       erspan_opt_policy, extack);
175 	if (err < 0)
176 		return err;
177 
178 	if (!tb[TCA_TUNNEL_KEY_ENC_OPT_ERSPAN_VER]) {
179 		NL_SET_ERR_MSG(extack, "Missing tunnel key erspan option ver");
180 		return -EINVAL;
181 	}
182 
183 	ver = nla_get_u8(tb[TCA_TUNNEL_KEY_ENC_OPT_ERSPAN_VER]);
184 	if (ver == 1) {
185 		if (!tb[TCA_TUNNEL_KEY_ENC_OPT_ERSPAN_INDEX]) {
186 			NL_SET_ERR_MSG(extack, "Missing tunnel key erspan option index");
187 			return -EINVAL;
188 		}
189 	} else if (ver == 2) {
190 		if (!tb[TCA_TUNNEL_KEY_ENC_OPT_ERSPAN_DIR] ||
191 		    !tb[TCA_TUNNEL_KEY_ENC_OPT_ERSPAN_HWID]) {
192 			NL_SET_ERR_MSG(extack, "Missing tunnel key erspan option dir or hwid");
193 			return -EINVAL;
194 		}
195 	} else {
196 		NL_SET_ERR_MSG(extack, "Tunnel key erspan option ver is incorrect");
197 		return -EINVAL;
198 	}
199 
200 	if (dst) {
201 		struct erspan_metadata *md = dst;
202 
203 		md->version = ver;
204 		if (ver == 1) {
205 			nla = tb[TCA_TUNNEL_KEY_ENC_OPT_ERSPAN_INDEX];
206 			md->u.index = nla_get_be32(nla);
207 		} else {
208 			nla = tb[TCA_TUNNEL_KEY_ENC_OPT_ERSPAN_DIR];
209 			md->u.md2.dir = nla_get_u8(nla);
210 			nla = tb[TCA_TUNNEL_KEY_ENC_OPT_ERSPAN_HWID];
211 			set_hwid(&md->u.md2, nla_get_u8(nla));
212 		}
213 	}
214 
215 	return sizeof(struct erspan_metadata);
216 }
217 
218 static int tunnel_key_copy_opts(const struct nlattr *nla, u8 *dst,
219 				int dst_len, struct netlink_ext_ack *extack)
220 {
221 	int err, rem, opt_len, len = nla_len(nla), opts_len = 0, type = 0;
222 	const struct nlattr *attr, *head = nla_data(nla);
223 
224 	err = nla_validate_deprecated(head, len, TCA_TUNNEL_KEY_ENC_OPTS_MAX,
225 				      enc_opts_policy, extack);
226 	if (err)
227 		return err;
228 
229 	nla_for_each_attr(attr, head, len, rem) {
230 		switch (nla_type(attr)) {
231 		case TCA_TUNNEL_KEY_ENC_OPTS_GENEVE:
232 			if (type && type != TUNNEL_GENEVE_OPT) {
233 				NL_SET_ERR_MSG(extack, "Duplicate type for geneve options");
234 				return -EINVAL;
235 			}
236 			opt_len = tunnel_key_copy_geneve_opt(attr, dst,
237 							     dst_len, extack);
238 			if (opt_len < 0)
239 				return opt_len;
240 			opts_len += opt_len;
241 			if (opts_len > IP_TUNNEL_OPTS_MAX) {
242 				NL_SET_ERR_MSG(extack, "Tunnel options exceeds max size");
243 				return -EINVAL;
244 			}
245 			if (dst) {
246 				dst_len -= opt_len;
247 				dst += opt_len;
248 			}
249 			type = TUNNEL_GENEVE_OPT;
250 			break;
251 		case TCA_TUNNEL_KEY_ENC_OPTS_VXLAN:
252 			if (type) {
253 				NL_SET_ERR_MSG(extack, "Duplicate type for vxlan options");
254 				return -EINVAL;
255 			}
256 			opt_len = tunnel_key_copy_vxlan_opt(attr, dst,
257 							    dst_len, extack);
258 			if (opt_len < 0)
259 				return opt_len;
260 			opts_len += opt_len;
261 			type = TUNNEL_VXLAN_OPT;
262 			break;
263 		case TCA_TUNNEL_KEY_ENC_OPTS_ERSPAN:
264 			if (type) {
265 				NL_SET_ERR_MSG(extack, "Duplicate type for erspan options");
266 				return -EINVAL;
267 			}
268 			opt_len = tunnel_key_copy_erspan_opt(attr, dst,
269 							     dst_len, extack);
270 			if (opt_len < 0)
271 				return opt_len;
272 			opts_len += opt_len;
273 			type = TUNNEL_ERSPAN_OPT;
274 			break;
275 		}
276 	}
277 
278 	if (!opts_len) {
279 		NL_SET_ERR_MSG(extack, "Empty list of tunnel options");
280 		return -EINVAL;
281 	}
282 
283 	if (rem > 0) {
284 		NL_SET_ERR_MSG(extack, "Trailing data after parsing tunnel key options attributes");
285 		return -EINVAL;
286 	}
287 
288 	return opts_len;
289 }
290 
291 static int tunnel_key_get_opts_len(struct nlattr *nla,
292 				   struct netlink_ext_ack *extack)
293 {
294 	return tunnel_key_copy_opts(nla, NULL, 0, extack);
295 }
296 
297 static int tunnel_key_opts_set(struct nlattr *nla, struct ip_tunnel_info *info,
298 			       int opts_len, struct netlink_ext_ack *extack)
299 {
300 	info->options_len = opts_len;
301 	switch (nla_type(nla_data(nla))) {
302 	case TCA_TUNNEL_KEY_ENC_OPTS_GENEVE:
303 #if IS_ENABLED(CONFIG_INET)
304 		info->key.tun_flags |= TUNNEL_GENEVE_OPT;
305 		return tunnel_key_copy_opts(nla, ip_tunnel_info_opts(info),
306 					    opts_len, extack);
307 #else
308 		return -EAFNOSUPPORT;
309 #endif
310 	case TCA_TUNNEL_KEY_ENC_OPTS_VXLAN:
311 #if IS_ENABLED(CONFIG_INET)
312 		info->key.tun_flags |= TUNNEL_VXLAN_OPT;
313 		return tunnel_key_copy_opts(nla, ip_tunnel_info_opts(info),
314 					    opts_len, extack);
315 #else
316 		return -EAFNOSUPPORT;
317 #endif
318 	case TCA_TUNNEL_KEY_ENC_OPTS_ERSPAN:
319 #if IS_ENABLED(CONFIG_INET)
320 		info->key.tun_flags |= TUNNEL_ERSPAN_OPT;
321 		return tunnel_key_copy_opts(nla, ip_tunnel_info_opts(info),
322 					    opts_len, extack);
323 #else
324 		return -EAFNOSUPPORT;
325 #endif
326 	default:
327 		NL_SET_ERR_MSG(extack, "Cannot set tunnel options for unknown tunnel type");
328 		return -EINVAL;
329 	}
330 }
331 
332 static const struct nla_policy tunnel_key_policy[TCA_TUNNEL_KEY_MAX + 1] = {
333 	[TCA_TUNNEL_KEY_PARMS]	    = { .len = sizeof(struct tc_tunnel_key) },
334 	[TCA_TUNNEL_KEY_ENC_IPV4_SRC] = { .type = NLA_U32 },
335 	[TCA_TUNNEL_KEY_ENC_IPV4_DST] = { .type = NLA_U32 },
336 	[TCA_TUNNEL_KEY_ENC_IPV6_SRC] = { .len = sizeof(struct in6_addr) },
337 	[TCA_TUNNEL_KEY_ENC_IPV6_DST] = { .len = sizeof(struct in6_addr) },
338 	[TCA_TUNNEL_KEY_ENC_KEY_ID]   = { .type = NLA_U32 },
339 	[TCA_TUNNEL_KEY_ENC_DST_PORT] = {.type = NLA_U16},
340 	[TCA_TUNNEL_KEY_NO_CSUM]      = { .type = NLA_U8 },
341 	[TCA_TUNNEL_KEY_ENC_OPTS]     = { .type = NLA_NESTED },
342 	[TCA_TUNNEL_KEY_ENC_TOS]      = { .type = NLA_U8 },
343 	[TCA_TUNNEL_KEY_ENC_TTL]      = { .type = NLA_U8 },
344 };
345 
346 static void tunnel_key_release_params(struct tcf_tunnel_key_params *p)
347 {
348 	if (!p)
349 		return;
350 	if (p->tcft_action == TCA_TUNNEL_KEY_ACT_SET)
351 		dst_release(&p->tcft_enc_metadata->dst);
352 
353 	kfree_rcu(p, rcu);
354 }
355 
356 static int tunnel_key_init(struct net *net, struct nlattr *nla,
357 			   struct nlattr *est, struct tc_action **a,
358 			   int ovr, int bind, bool rtnl_held,
359 			   struct tcf_proto *tp, u32 act_flags,
360 			   struct netlink_ext_ack *extack)
361 {
362 	struct tc_action_net *tn = net_generic(net, tunnel_key_net_id);
363 	struct nlattr *tb[TCA_TUNNEL_KEY_MAX + 1];
364 	struct tcf_tunnel_key_params *params_new;
365 	struct metadata_dst *metadata = NULL;
366 	struct tcf_chain *goto_ch = NULL;
367 	struct tc_tunnel_key *parm;
368 	struct tcf_tunnel_key *t;
369 	bool exists = false;
370 	__be16 dst_port = 0;
371 	__be64 key_id = 0;
372 	int opts_len = 0;
373 	__be16 flags = 0;
374 	u8 tos, ttl;
375 	int ret = 0;
376 	u32 index;
377 	int err;
378 
379 	if (!nla) {
380 		NL_SET_ERR_MSG(extack, "Tunnel requires attributes to be passed");
381 		return -EINVAL;
382 	}
383 
384 	err = nla_parse_nested_deprecated(tb, TCA_TUNNEL_KEY_MAX, nla,
385 					  tunnel_key_policy, extack);
386 	if (err < 0) {
387 		NL_SET_ERR_MSG(extack, "Failed to parse nested tunnel key attributes");
388 		return err;
389 	}
390 
391 	if (!tb[TCA_TUNNEL_KEY_PARMS]) {
392 		NL_SET_ERR_MSG(extack, "Missing tunnel key parameters");
393 		return -EINVAL;
394 	}
395 
396 	parm = nla_data(tb[TCA_TUNNEL_KEY_PARMS]);
397 	index = parm->index;
398 	err = tcf_idr_check_alloc(tn, &index, a, bind);
399 	if (err < 0)
400 		return err;
401 	exists = err;
402 	if (exists && bind)
403 		return 0;
404 
405 	switch (parm->t_action) {
406 	case TCA_TUNNEL_KEY_ACT_RELEASE:
407 		break;
408 	case TCA_TUNNEL_KEY_ACT_SET:
409 		if (tb[TCA_TUNNEL_KEY_ENC_KEY_ID]) {
410 			__be32 key32;
411 
412 			key32 = nla_get_be32(tb[TCA_TUNNEL_KEY_ENC_KEY_ID]);
413 			key_id = key32_to_tunnel_id(key32);
414 			flags = TUNNEL_KEY;
415 		}
416 
417 		flags |= TUNNEL_CSUM;
418 		if (tb[TCA_TUNNEL_KEY_NO_CSUM] &&
419 		    nla_get_u8(tb[TCA_TUNNEL_KEY_NO_CSUM]))
420 			flags &= ~TUNNEL_CSUM;
421 
422 		if (tb[TCA_TUNNEL_KEY_ENC_DST_PORT])
423 			dst_port = nla_get_be16(tb[TCA_TUNNEL_KEY_ENC_DST_PORT]);
424 
425 		if (tb[TCA_TUNNEL_KEY_ENC_OPTS]) {
426 			opts_len = tunnel_key_get_opts_len(tb[TCA_TUNNEL_KEY_ENC_OPTS],
427 							   extack);
428 			if (opts_len < 0) {
429 				ret = opts_len;
430 				goto err_out;
431 			}
432 		}
433 
434 		tos = 0;
435 		if (tb[TCA_TUNNEL_KEY_ENC_TOS])
436 			tos = nla_get_u8(tb[TCA_TUNNEL_KEY_ENC_TOS]);
437 		ttl = 0;
438 		if (tb[TCA_TUNNEL_KEY_ENC_TTL])
439 			ttl = nla_get_u8(tb[TCA_TUNNEL_KEY_ENC_TTL]);
440 
441 		if (tb[TCA_TUNNEL_KEY_ENC_IPV4_SRC] &&
442 		    tb[TCA_TUNNEL_KEY_ENC_IPV4_DST]) {
443 			__be32 saddr;
444 			__be32 daddr;
445 
446 			saddr = nla_get_in_addr(tb[TCA_TUNNEL_KEY_ENC_IPV4_SRC]);
447 			daddr = nla_get_in_addr(tb[TCA_TUNNEL_KEY_ENC_IPV4_DST]);
448 
449 			metadata = __ip_tun_set_dst(saddr, daddr, tos, ttl,
450 						    dst_port, flags,
451 						    key_id, opts_len);
452 		} else if (tb[TCA_TUNNEL_KEY_ENC_IPV6_SRC] &&
453 			   tb[TCA_TUNNEL_KEY_ENC_IPV6_DST]) {
454 			struct in6_addr saddr;
455 			struct in6_addr daddr;
456 
457 			saddr = nla_get_in6_addr(tb[TCA_TUNNEL_KEY_ENC_IPV6_SRC]);
458 			daddr = nla_get_in6_addr(tb[TCA_TUNNEL_KEY_ENC_IPV6_DST]);
459 
460 			metadata = __ipv6_tun_set_dst(&saddr, &daddr, tos, ttl, dst_port,
461 						      0, flags,
462 						      key_id, opts_len);
463 		} else {
464 			NL_SET_ERR_MSG(extack, "Missing either ipv4 or ipv6 src and dst");
465 			ret = -EINVAL;
466 			goto err_out;
467 		}
468 
469 		if (!metadata) {
470 			NL_SET_ERR_MSG(extack, "Cannot allocate tunnel metadata dst");
471 			ret = -ENOMEM;
472 			goto err_out;
473 		}
474 
475 #ifdef CONFIG_DST_CACHE
476 		ret = dst_cache_init(&metadata->u.tun_info.dst_cache, GFP_KERNEL);
477 		if (ret)
478 			goto release_tun_meta;
479 #endif
480 
481 		if (opts_len) {
482 			ret = tunnel_key_opts_set(tb[TCA_TUNNEL_KEY_ENC_OPTS],
483 						  &metadata->u.tun_info,
484 						  opts_len, extack);
485 			if (ret < 0)
486 				goto release_tun_meta;
487 		}
488 
489 		metadata->u.tun_info.mode |= IP_TUNNEL_INFO_TX;
490 		break;
491 	default:
492 		NL_SET_ERR_MSG(extack, "Unknown tunnel key action");
493 		ret = -EINVAL;
494 		goto err_out;
495 	}
496 
497 	if (!exists) {
498 		ret = tcf_idr_create_from_flags(tn, index, est, a,
499 						&act_tunnel_key_ops, bind,
500 						act_flags);
501 		if (ret) {
502 			NL_SET_ERR_MSG(extack, "Cannot create TC IDR");
503 			goto release_tun_meta;
504 		}
505 
506 		ret = ACT_P_CREATED;
507 	} else if (!ovr) {
508 		NL_SET_ERR_MSG(extack, "TC IDR already exists");
509 		ret = -EEXIST;
510 		goto release_tun_meta;
511 	}
512 
513 	err = tcf_action_check_ctrlact(parm->action, tp, &goto_ch, extack);
514 	if (err < 0) {
515 		ret = err;
516 		exists = true;
517 		goto release_tun_meta;
518 	}
519 	t = to_tunnel_key(*a);
520 
521 	params_new = kzalloc(sizeof(*params_new), GFP_KERNEL);
522 	if (unlikely(!params_new)) {
523 		NL_SET_ERR_MSG(extack, "Cannot allocate tunnel key parameters");
524 		ret = -ENOMEM;
525 		exists = true;
526 		goto put_chain;
527 	}
528 	params_new->tcft_action = parm->t_action;
529 	params_new->tcft_enc_metadata = metadata;
530 
531 	spin_lock_bh(&t->tcf_lock);
532 	goto_ch = tcf_action_set_ctrlact(*a, parm->action, goto_ch);
533 	params_new = rcu_replace_pointer(t->params, params_new,
534 					 lockdep_is_held(&t->tcf_lock));
535 	spin_unlock_bh(&t->tcf_lock);
536 	tunnel_key_release_params(params_new);
537 	if (goto_ch)
538 		tcf_chain_put_by_act(goto_ch);
539 
540 	return ret;
541 
542 put_chain:
543 	if (goto_ch)
544 		tcf_chain_put_by_act(goto_ch);
545 
546 release_tun_meta:
547 	if (metadata)
548 		dst_release(&metadata->dst);
549 
550 err_out:
551 	if (exists)
552 		tcf_idr_release(*a, bind);
553 	else
554 		tcf_idr_cleanup(tn, index);
555 	return ret;
556 }
557 
558 static void tunnel_key_release(struct tc_action *a)
559 {
560 	struct tcf_tunnel_key *t = to_tunnel_key(a);
561 	struct tcf_tunnel_key_params *params;
562 
563 	params = rcu_dereference_protected(t->params, 1);
564 	tunnel_key_release_params(params);
565 }
566 
567 static int tunnel_key_geneve_opts_dump(struct sk_buff *skb,
568 				       const struct ip_tunnel_info *info)
569 {
570 	int len = info->options_len;
571 	u8 *src = (u8 *)(info + 1);
572 	struct nlattr *start;
573 
574 	start = nla_nest_start_noflag(skb, TCA_TUNNEL_KEY_ENC_OPTS_GENEVE);
575 	if (!start)
576 		return -EMSGSIZE;
577 
578 	while (len > 0) {
579 		struct geneve_opt *opt = (struct geneve_opt *)src;
580 
581 		if (nla_put_be16(skb, TCA_TUNNEL_KEY_ENC_OPT_GENEVE_CLASS,
582 				 opt->opt_class) ||
583 		    nla_put_u8(skb, TCA_TUNNEL_KEY_ENC_OPT_GENEVE_TYPE,
584 			       opt->type) ||
585 		    nla_put(skb, TCA_TUNNEL_KEY_ENC_OPT_GENEVE_DATA,
586 			    opt->length * 4, opt + 1)) {
587 			nla_nest_cancel(skb, start);
588 			return -EMSGSIZE;
589 		}
590 
591 		len -= sizeof(struct geneve_opt) + opt->length * 4;
592 		src += sizeof(struct geneve_opt) + opt->length * 4;
593 	}
594 
595 	nla_nest_end(skb, start);
596 	return 0;
597 }
598 
599 static int tunnel_key_vxlan_opts_dump(struct sk_buff *skb,
600 				      const struct ip_tunnel_info *info)
601 {
602 	struct vxlan_metadata *md = (struct vxlan_metadata *)(info + 1);
603 	struct nlattr *start;
604 
605 	start = nla_nest_start_noflag(skb, TCA_TUNNEL_KEY_ENC_OPTS_VXLAN);
606 	if (!start)
607 		return -EMSGSIZE;
608 
609 	if (nla_put_u32(skb, TCA_TUNNEL_KEY_ENC_OPT_VXLAN_GBP, md->gbp)) {
610 		nla_nest_cancel(skb, start);
611 		return -EMSGSIZE;
612 	}
613 
614 	nla_nest_end(skb, start);
615 	return 0;
616 }
617 
618 static int tunnel_key_erspan_opts_dump(struct sk_buff *skb,
619 				       const struct ip_tunnel_info *info)
620 {
621 	struct erspan_metadata *md = (struct erspan_metadata *)(info + 1);
622 	struct nlattr *start;
623 
624 	start = nla_nest_start_noflag(skb, TCA_TUNNEL_KEY_ENC_OPTS_ERSPAN);
625 	if (!start)
626 		return -EMSGSIZE;
627 
628 	if (nla_put_u8(skb, TCA_TUNNEL_KEY_ENC_OPT_ERSPAN_VER, md->version))
629 		goto err;
630 
631 	if (md->version == 1 &&
632 	    nla_put_be32(skb, TCA_TUNNEL_KEY_ENC_OPT_ERSPAN_INDEX, md->u.index))
633 		goto err;
634 
635 	if (md->version == 2 &&
636 	    (nla_put_u8(skb, TCA_TUNNEL_KEY_ENC_OPT_ERSPAN_DIR,
637 			md->u.md2.dir) ||
638 	     nla_put_u8(skb, TCA_TUNNEL_KEY_ENC_OPT_ERSPAN_HWID,
639 			get_hwid(&md->u.md2))))
640 		goto err;
641 
642 	nla_nest_end(skb, start);
643 	return 0;
644 err:
645 	nla_nest_cancel(skb, start);
646 	return -EMSGSIZE;
647 }
648 
649 static int tunnel_key_opts_dump(struct sk_buff *skb,
650 				const struct ip_tunnel_info *info)
651 {
652 	struct nlattr *start;
653 	int err = -EINVAL;
654 
655 	if (!info->options_len)
656 		return 0;
657 
658 	start = nla_nest_start_noflag(skb, TCA_TUNNEL_KEY_ENC_OPTS);
659 	if (!start)
660 		return -EMSGSIZE;
661 
662 	if (info->key.tun_flags & TUNNEL_GENEVE_OPT) {
663 		err = tunnel_key_geneve_opts_dump(skb, info);
664 		if (err)
665 			goto err_out;
666 	} else if (info->key.tun_flags & TUNNEL_VXLAN_OPT) {
667 		err = tunnel_key_vxlan_opts_dump(skb, info);
668 		if (err)
669 			goto err_out;
670 	} else if (info->key.tun_flags & TUNNEL_ERSPAN_OPT) {
671 		err = tunnel_key_erspan_opts_dump(skb, info);
672 		if (err)
673 			goto err_out;
674 	} else {
675 err_out:
676 		nla_nest_cancel(skb, start);
677 		return err;
678 	}
679 
680 	nla_nest_end(skb, start);
681 	return 0;
682 }
683 
684 static int tunnel_key_dump_addresses(struct sk_buff *skb,
685 				     const struct ip_tunnel_info *info)
686 {
687 	unsigned short family = ip_tunnel_info_af(info);
688 
689 	if (family == AF_INET) {
690 		__be32 saddr = info->key.u.ipv4.src;
691 		__be32 daddr = info->key.u.ipv4.dst;
692 
693 		if (!nla_put_in_addr(skb, TCA_TUNNEL_KEY_ENC_IPV4_SRC, saddr) &&
694 		    !nla_put_in_addr(skb, TCA_TUNNEL_KEY_ENC_IPV4_DST, daddr))
695 			return 0;
696 	}
697 
698 	if (family == AF_INET6) {
699 		const struct in6_addr *saddr6 = &info->key.u.ipv6.src;
700 		const struct in6_addr *daddr6 = &info->key.u.ipv6.dst;
701 
702 		if (!nla_put_in6_addr(skb,
703 				      TCA_TUNNEL_KEY_ENC_IPV6_SRC, saddr6) &&
704 		    !nla_put_in6_addr(skb,
705 				      TCA_TUNNEL_KEY_ENC_IPV6_DST, daddr6))
706 			return 0;
707 	}
708 
709 	return -EINVAL;
710 }
711 
712 static int tunnel_key_dump(struct sk_buff *skb, struct tc_action *a,
713 			   int bind, int ref)
714 {
715 	unsigned char *b = skb_tail_pointer(skb);
716 	struct tcf_tunnel_key *t = to_tunnel_key(a);
717 	struct tcf_tunnel_key_params *params;
718 	struct tc_tunnel_key opt = {
719 		.index    = t->tcf_index,
720 		.refcnt   = refcount_read(&t->tcf_refcnt) - ref,
721 		.bindcnt  = atomic_read(&t->tcf_bindcnt) - bind,
722 	};
723 	struct tcf_t tm;
724 
725 	spin_lock_bh(&t->tcf_lock);
726 	params = rcu_dereference_protected(t->params,
727 					   lockdep_is_held(&t->tcf_lock));
728 	opt.action   = t->tcf_action;
729 	opt.t_action = params->tcft_action;
730 
731 	if (nla_put(skb, TCA_TUNNEL_KEY_PARMS, sizeof(opt), &opt))
732 		goto nla_put_failure;
733 
734 	if (params->tcft_action == TCA_TUNNEL_KEY_ACT_SET) {
735 		struct ip_tunnel_info *info =
736 			&params->tcft_enc_metadata->u.tun_info;
737 		struct ip_tunnel_key *key = &info->key;
738 		__be32 key_id = tunnel_id_to_key32(key->tun_id);
739 
740 		if (((key->tun_flags & TUNNEL_KEY) &&
741 		     nla_put_be32(skb, TCA_TUNNEL_KEY_ENC_KEY_ID, key_id)) ||
742 		    tunnel_key_dump_addresses(skb,
743 					      &params->tcft_enc_metadata->u.tun_info) ||
744 		    (key->tp_dst &&
745 		      nla_put_be16(skb, TCA_TUNNEL_KEY_ENC_DST_PORT,
746 				   key->tp_dst)) ||
747 		    nla_put_u8(skb, TCA_TUNNEL_KEY_NO_CSUM,
748 			       !(key->tun_flags & TUNNEL_CSUM)) ||
749 		    tunnel_key_opts_dump(skb, info))
750 			goto nla_put_failure;
751 
752 		if (key->tos && nla_put_u8(skb, TCA_TUNNEL_KEY_ENC_TOS, key->tos))
753 			goto nla_put_failure;
754 
755 		if (key->ttl && nla_put_u8(skb, TCA_TUNNEL_KEY_ENC_TTL, key->ttl))
756 			goto nla_put_failure;
757 	}
758 
759 	tcf_tm_dump(&tm, &t->tcf_tm);
760 	if (nla_put_64bit(skb, TCA_TUNNEL_KEY_TM, sizeof(tm),
761 			  &tm, TCA_TUNNEL_KEY_PAD))
762 		goto nla_put_failure;
763 	spin_unlock_bh(&t->tcf_lock);
764 
765 	return skb->len;
766 
767 nla_put_failure:
768 	spin_unlock_bh(&t->tcf_lock);
769 	nlmsg_trim(skb, b);
770 	return -1;
771 }
772 
773 static int tunnel_key_walker(struct net *net, struct sk_buff *skb,
774 			     struct netlink_callback *cb, int type,
775 			     const struct tc_action_ops *ops,
776 			     struct netlink_ext_ack *extack)
777 {
778 	struct tc_action_net *tn = net_generic(net, tunnel_key_net_id);
779 
780 	return tcf_generic_walker(tn, skb, cb, type, ops, extack);
781 }
782 
783 static int tunnel_key_search(struct net *net, struct tc_action **a, u32 index)
784 {
785 	struct tc_action_net *tn = net_generic(net, tunnel_key_net_id);
786 
787 	return tcf_idr_search(tn, a, index);
788 }
789 
790 static struct tc_action_ops act_tunnel_key_ops = {
791 	.kind		=	"tunnel_key",
792 	.id		=	TCA_ID_TUNNEL_KEY,
793 	.owner		=	THIS_MODULE,
794 	.act		=	tunnel_key_act,
795 	.dump		=	tunnel_key_dump,
796 	.init		=	tunnel_key_init,
797 	.cleanup	=	tunnel_key_release,
798 	.walk		=	tunnel_key_walker,
799 	.lookup		=	tunnel_key_search,
800 	.size		=	sizeof(struct tcf_tunnel_key),
801 };
802 
803 static __net_init int tunnel_key_init_net(struct net *net)
804 {
805 	struct tc_action_net *tn = net_generic(net, tunnel_key_net_id);
806 
807 	return tc_action_net_init(net, tn, &act_tunnel_key_ops);
808 }
809 
810 static void __net_exit tunnel_key_exit_net(struct list_head *net_list)
811 {
812 	tc_action_net_exit(net_list, tunnel_key_net_id);
813 }
814 
815 static struct pernet_operations tunnel_key_net_ops = {
816 	.init = tunnel_key_init_net,
817 	.exit_batch = tunnel_key_exit_net,
818 	.id   = &tunnel_key_net_id,
819 	.size = sizeof(struct tc_action_net),
820 };
821 
822 static int __init tunnel_key_init_module(void)
823 {
824 	return tcf_register_action(&act_tunnel_key_ops, &tunnel_key_net_ops);
825 }
826 
827 static void __exit tunnel_key_cleanup_module(void)
828 {
829 	tcf_unregister_action(&act_tunnel_key_ops, &tunnel_key_net_ops);
830 }
831 
832 module_init(tunnel_key_init_module);
833 module_exit(tunnel_key_cleanup_module);
834 
835 MODULE_AUTHOR("Amir Vadai <amir@vadai.me>");
836 MODULE_DESCRIPTION("ip tunnel manipulation actions");
837 MODULE_LICENSE("GPL v2");
838