1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * Copyright (c) 2016, Amir Vadai <amir@vadai.me>
4 * Copyright (c) 2016, Mellanox Technologies. All rights reserved.
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/netlink.h>
14 #include <net/pkt_sched.h>
16 #include <net/pkt_cls.h>
18 #include <linux/tc_act/tc_tunnel_key.h>
19 #include <net/tc_act/tc_tunnel_key.h>
21 static unsigned int tunnel_key_net_id;
22 static struct tc_action_ops act_tunnel_key_ops;
24 static int tunnel_key_act(struct sk_buff *skb, const struct tc_action *a,
25 struct tcf_result *res)
27 struct tcf_tunnel_key *t = to_tunnel_key(a);
28 struct tcf_tunnel_key_params *params;
31 params = rcu_dereference_bh(t->params);
33 tcf_lastuse_update(&t->tcf_tm);
34 bstats_cpu_update(this_cpu_ptr(t->common.cpu_bstats), skb);
35 action = READ_ONCE(t->tcf_action);
37 switch (params->tcft_action) {
38 case TCA_TUNNEL_KEY_ACT_RELEASE:
41 case TCA_TUNNEL_KEY_ACT_SET:
43 skb_dst_set(skb, dst_clone(¶ms->tcft_enc_metadata->dst));
46 WARN_ONCE(1, "Bad tunnel_key action %d.\n",
54 static const struct nla_policy
55 enc_opts_policy[TCA_TUNNEL_KEY_ENC_OPTS_MAX + 1] = {
56 [TCA_TUNNEL_KEY_ENC_OPTS_GENEVE] = { .type = NLA_NESTED },
59 static const struct nla_policy
60 geneve_opt_policy[TCA_TUNNEL_KEY_ENC_OPT_GENEVE_MAX + 1] = {
61 [TCA_TUNNEL_KEY_ENC_OPT_GENEVE_CLASS] = { .type = NLA_U16 },
62 [TCA_TUNNEL_KEY_ENC_OPT_GENEVE_TYPE] = { .type = NLA_U8 },
63 [TCA_TUNNEL_KEY_ENC_OPT_GENEVE_DATA] = { .type = NLA_BINARY,
68 tunnel_key_copy_geneve_opt(const struct nlattr *nla, void *dst, int dst_len,
69 struct netlink_ext_ack *extack)
71 struct nlattr *tb[TCA_TUNNEL_KEY_ENC_OPT_GENEVE_MAX + 1];
72 int err, data_len, opt_len;
75 err = nla_parse_nested_deprecated(tb,
76 TCA_TUNNEL_KEY_ENC_OPT_GENEVE_MAX,
77 nla, geneve_opt_policy, extack);
81 if (!tb[TCA_TUNNEL_KEY_ENC_OPT_GENEVE_CLASS] ||
82 !tb[TCA_TUNNEL_KEY_ENC_OPT_GENEVE_TYPE] ||
83 !tb[TCA_TUNNEL_KEY_ENC_OPT_GENEVE_DATA]) {
84 NL_SET_ERR_MSG(extack, "Missing tunnel key geneve option class, type or data");
88 data = nla_data(tb[TCA_TUNNEL_KEY_ENC_OPT_GENEVE_DATA]);
89 data_len = nla_len(tb[TCA_TUNNEL_KEY_ENC_OPT_GENEVE_DATA]);
91 NL_SET_ERR_MSG(extack, "Tunnel key geneve option data is less than 4 bytes long");
95 NL_SET_ERR_MSG(extack, "Tunnel key geneve option data is not a multiple of 4 bytes long");
99 opt_len = sizeof(struct geneve_opt) + data_len;
101 struct geneve_opt *opt = dst;
103 WARN_ON(dst_len < opt_len);
106 nla_get_be16(tb[TCA_TUNNEL_KEY_ENC_OPT_GENEVE_CLASS]);
107 opt->type = nla_get_u8(tb[TCA_TUNNEL_KEY_ENC_OPT_GENEVE_TYPE]);
108 opt->length = data_len / 4; /* length is in units of 4 bytes */
113 memcpy(opt + 1, data, data_len);
119 static int tunnel_key_copy_opts(const struct nlattr *nla, u8 *dst,
120 int dst_len, struct netlink_ext_ack *extack)
122 int err, rem, opt_len, len = nla_len(nla), opts_len = 0;
123 const struct nlattr *attr, *head = nla_data(nla);
125 err = nla_validate_deprecated(head, len, TCA_TUNNEL_KEY_ENC_OPTS_MAX,
126 enc_opts_policy, extack);
130 nla_for_each_attr(attr, head, len, rem) {
131 switch (nla_type(attr)) {
132 case TCA_TUNNEL_KEY_ENC_OPTS_GENEVE:
133 opt_len = tunnel_key_copy_geneve_opt(attr, dst,
147 NL_SET_ERR_MSG(extack, "Empty list of tunnel options");
152 NL_SET_ERR_MSG(extack, "Trailing data after parsing tunnel key options attributes");
159 static int tunnel_key_get_opts_len(struct nlattr *nla,
160 struct netlink_ext_ack *extack)
162 return tunnel_key_copy_opts(nla, NULL, 0, extack);
165 static int tunnel_key_opts_set(struct nlattr *nla, struct ip_tunnel_info *info,
166 int opts_len, struct netlink_ext_ack *extack)
168 info->options_len = opts_len;
169 switch (nla_type(nla_data(nla))) {
170 case TCA_TUNNEL_KEY_ENC_OPTS_GENEVE:
171 #if IS_ENABLED(CONFIG_INET)
172 info->key.tun_flags |= TUNNEL_GENEVE_OPT;
173 return tunnel_key_copy_opts(nla, ip_tunnel_info_opts(info),
176 return -EAFNOSUPPORT;
179 NL_SET_ERR_MSG(extack, "Cannot set tunnel options for unknown tunnel type");
184 static const struct nla_policy tunnel_key_policy[TCA_TUNNEL_KEY_MAX + 1] = {
185 [TCA_TUNNEL_KEY_PARMS] = { .len = sizeof(struct tc_tunnel_key) },
186 [TCA_TUNNEL_KEY_ENC_IPV4_SRC] = { .type = NLA_U32 },
187 [TCA_TUNNEL_KEY_ENC_IPV4_DST] = { .type = NLA_U32 },
188 [TCA_TUNNEL_KEY_ENC_IPV6_SRC] = { .len = sizeof(struct in6_addr) },
189 [TCA_TUNNEL_KEY_ENC_IPV6_DST] = { .len = sizeof(struct in6_addr) },
190 [TCA_TUNNEL_KEY_ENC_KEY_ID] = { .type = NLA_U32 },
191 [TCA_TUNNEL_KEY_ENC_DST_PORT] = {.type = NLA_U16},
192 [TCA_TUNNEL_KEY_NO_CSUM] = { .type = NLA_U8 },
193 [TCA_TUNNEL_KEY_ENC_OPTS] = { .type = NLA_NESTED },
194 [TCA_TUNNEL_KEY_ENC_TOS] = { .type = NLA_U8 },
195 [TCA_TUNNEL_KEY_ENC_TTL] = { .type = NLA_U8 },
198 static void tunnel_key_release_params(struct tcf_tunnel_key_params *p)
202 if (p->tcft_action == TCA_TUNNEL_KEY_ACT_SET)
203 dst_release(&p->tcft_enc_metadata->dst);
208 static int tunnel_key_init(struct net *net, struct nlattr *nla,
209 struct nlattr *est, struct tc_action **a,
210 int ovr, int bind, bool rtnl_held,
211 struct tcf_proto *tp,
212 struct netlink_ext_ack *extack)
214 struct tc_action_net *tn = net_generic(net, tunnel_key_net_id);
215 struct nlattr *tb[TCA_TUNNEL_KEY_MAX + 1];
216 struct tcf_tunnel_key_params *params_new;
217 struct metadata_dst *metadata = NULL;
218 struct tcf_chain *goto_ch = NULL;
219 struct tc_tunnel_key *parm;
220 struct tcf_tunnel_key *t;
231 NL_SET_ERR_MSG(extack, "Tunnel requires attributes to be passed");
235 err = nla_parse_nested_deprecated(tb, TCA_TUNNEL_KEY_MAX, nla,
236 tunnel_key_policy, extack);
238 NL_SET_ERR_MSG(extack, "Failed to parse nested tunnel key attributes");
242 if (!tb[TCA_TUNNEL_KEY_PARMS]) {
243 NL_SET_ERR_MSG(extack, "Missing tunnel key parameters");
247 parm = nla_data(tb[TCA_TUNNEL_KEY_PARMS]);
248 err = tcf_idr_check_alloc(tn, &parm->index, a, bind);
255 switch (parm->t_action) {
256 case TCA_TUNNEL_KEY_ACT_RELEASE:
258 case TCA_TUNNEL_KEY_ACT_SET:
259 if (tb[TCA_TUNNEL_KEY_ENC_KEY_ID]) {
262 key32 = nla_get_be32(tb[TCA_TUNNEL_KEY_ENC_KEY_ID]);
263 key_id = key32_to_tunnel_id(key32);
267 flags |= TUNNEL_CSUM;
268 if (tb[TCA_TUNNEL_KEY_NO_CSUM] &&
269 nla_get_u8(tb[TCA_TUNNEL_KEY_NO_CSUM]))
270 flags &= ~TUNNEL_CSUM;
272 if (tb[TCA_TUNNEL_KEY_ENC_DST_PORT])
273 dst_port = nla_get_be16(tb[TCA_TUNNEL_KEY_ENC_DST_PORT]);
275 if (tb[TCA_TUNNEL_KEY_ENC_OPTS]) {
276 opts_len = tunnel_key_get_opts_len(tb[TCA_TUNNEL_KEY_ENC_OPTS],
285 if (tb[TCA_TUNNEL_KEY_ENC_TOS])
286 tos = nla_get_u8(tb[TCA_TUNNEL_KEY_ENC_TOS]);
288 if (tb[TCA_TUNNEL_KEY_ENC_TTL])
289 ttl = nla_get_u8(tb[TCA_TUNNEL_KEY_ENC_TTL]);
291 if (tb[TCA_TUNNEL_KEY_ENC_IPV4_SRC] &&
292 tb[TCA_TUNNEL_KEY_ENC_IPV4_DST]) {
296 saddr = nla_get_in_addr(tb[TCA_TUNNEL_KEY_ENC_IPV4_SRC]);
297 daddr = nla_get_in_addr(tb[TCA_TUNNEL_KEY_ENC_IPV4_DST]);
299 metadata = __ip_tun_set_dst(saddr, daddr, tos, ttl,
302 } else if (tb[TCA_TUNNEL_KEY_ENC_IPV6_SRC] &&
303 tb[TCA_TUNNEL_KEY_ENC_IPV6_DST]) {
304 struct in6_addr saddr;
305 struct in6_addr daddr;
307 saddr = nla_get_in6_addr(tb[TCA_TUNNEL_KEY_ENC_IPV6_SRC]);
308 daddr = nla_get_in6_addr(tb[TCA_TUNNEL_KEY_ENC_IPV6_DST]);
310 metadata = __ipv6_tun_set_dst(&saddr, &daddr, tos, ttl, dst_port,
314 NL_SET_ERR_MSG(extack, "Missing either ipv4 or ipv6 src and dst");
320 NL_SET_ERR_MSG(extack, "Cannot allocate tunnel metadata dst");
325 #ifdef CONFIG_DST_CACHE
326 ret = dst_cache_init(&metadata->u.tun_info.dst_cache, GFP_KERNEL);
328 goto release_tun_meta;
332 ret = tunnel_key_opts_set(tb[TCA_TUNNEL_KEY_ENC_OPTS],
333 &metadata->u.tun_info,
336 goto release_tun_meta;
339 metadata->u.tun_info.mode |= IP_TUNNEL_INFO_TX;
342 NL_SET_ERR_MSG(extack, "Unknown tunnel key action");
348 ret = tcf_idr_create(tn, parm->index, est, a,
349 &act_tunnel_key_ops, bind, true);
351 NL_SET_ERR_MSG(extack, "Cannot create TC IDR");
352 goto release_tun_meta;
357 NL_SET_ERR_MSG(extack, "TC IDR already exists");
359 goto release_tun_meta;
362 err = tcf_action_check_ctrlact(parm->action, tp, &goto_ch, extack);
366 goto release_tun_meta;
368 t = to_tunnel_key(*a);
370 params_new = kzalloc(sizeof(*params_new), GFP_KERNEL);
371 if (unlikely(!params_new)) {
372 NL_SET_ERR_MSG(extack, "Cannot allocate tunnel key parameters");
377 params_new->tcft_action = parm->t_action;
378 params_new->tcft_enc_metadata = metadata;
380 spin_lock_bh(&t->tcf_lock);
381 goto_ch = tcf_action_set_ctrlact(*a, parm->action, goto_ch);
382 rcu_swap_protected(t->params, params_new,
383 lockdep_is_held(&t->tcf_lock));
384 spin_unlock_bh(&t->tcf_lock);
385 tunnel_key_release_params(params_new);
387 tcf_chain_put_by_act(goto_ch);
389 if (ret == ACT_P_CREATED)
390 tcf_idr_insert(tn, *a);
396 tcf_chain_put_by_act(goto_ch);
400 dst_release(&metadata->dst);
404 tcf_idr_release(*a, bind);
406 tcf_idr_cleanup(tn, parm->index);
410 static void tunnel_key_release(struct tc_action *a)
412 struct tcf_tunnel_key *t = to_tunnel_key(a);
413 struct tcf_tunnel_key_params *params;
415 params = rcu_dereference_protected(t->params, 1);
416 tunnel_key_release_params(params);
419 static int tunnel_key_geneve_opts_dump(struct sk_buff *skb,
420 const struct ip_tunnel_info *info)
422 int len = info->options_len;
423 u8 *src = (u8 *)(info + 1);
424 struct nlattr *start;
426 start = nla_nest_start_noflag(skb, TCA_TUNNEL_KEY_ENC_OPTS_GENEVE);
431 struct geneve_opt *opt = (struct geneve_opt *)src;
433 if (nla_put_be16(skb, TCA_TUNNEL_KEY_ENC_OPT_GENEVE_CLASS,
435 nla_put_u8(skb, TCA_TUNNEL_KEY_ENC_OPT_GENEVE_TYPE,
437 nla_put(skb, TCA_TUNNEL_KEY_ENC_OPT_GENEVE_DATA,
438 opt->length * 4, opt + 1)) {
439 nla_nest_cancel(skb, start);
443 len -= sizeof(struct geneve_opt) + opt->length * 4;
444 src += sizeof(struct geneve_opt) + opt->length * 4;
447 nla_nest_end(skb, start);
451 static int tunnel_key_opts_dump(struct sk_buff *skb,
452 const struct ip_tunnel_info *info)
454 struct nlattr *start;
457 if (!info->options_len)
460 start = nla_nest_start_noflag(skb, TCA_TUNNEL_KEY_ENC_OPTS);
464 if (info->key.tun_flags & TUNNEL_GENEVE_OPT) {
465 err = tunnel_key_geneve_opts_dump(skb, info);
470 nla_nest_cancel(skb, start);
474 nla_nest_end(skb, start);
478 static int tunnel_key_dump_addresses(struct sk_buff *skb,
479 const struct ip_tunnel_info *info)
481 unsigned short family = ip_tunnel_info_af(info);
483 if (family == AF_INET) {
484 __be32 saddr = info->key.u.ipv4.src;
485 __be32 daddr = info->key.u.ipv4.dst;
487 if (!nla_put_in_addr(skb, TCA_TUNNEL_KEY_ENC_IPV4_SRC, saddr) &&
488 !nla_put_in_addr(skb, TCA_TUNNEL_KEY_ENC_IPV4_DST, daddr))
492 if (family == AF_INET6) {
493 const struct in6_addr *saddr6 = &info->key.u.ipv6.src;
494 const struct in6_addr *daddr6 = &info->key.u.ipv6.dst;
496 if (!nla_put_in6_addr(skb,
497 TCA_TUNNEL_KEY_ENC_IPV6_SRC, saddr6) &&
498 !nla_put_in6_addr(skb,
499 TCA_TUNNEL_KEY_ENC_IPV6_DST, daddr6))
506 static int tunnel_key_dump(struct sk_buff *skb, struct tc_action *a,
509 unsigned char *b = skb_tail_pointer(skb);
510 struct tcf_tunnel_key *t = to_tunnel_key(a);
511 struct tcf_tunnel_key_params *params;
512 struct tc_tunnel_key opt = {
513 .index = t->tcf_index,
514 .refcnt = refcount_read(&t->tcf_refcnt) - ref,
515 .bindcnt = atomic_read(&t->tcf_bindcnt) - bind,
519 spin_lock_bh(&t->tcf_lock);
520 params = rcu_dereference_protected(t->params,
521 lockdep_is_held(&t->tcf_lock));
522 opt.action = t->tcf_action;
523 opt.t_action = params->tcft_action;
525 if (nla_put(skb, TCA_TUNNEL_KEY_PARMS, sizeof(opt), &opt))
526 goto nla_put_failure;
528 if (params->tcft_action == TCA_TUNNEL_KEY_ACT_SET) {
529 struct ip_tunnel_info *info =
530 ¶ms->tcft_enc_metadata->u.tun_info;
531 struct ip_tunnel_key *key = &info->key;
532 __be32 key_id = tunnel_id_to_key32(key->tun_id);
534 if (((key->tun_flags & TUNNEL_KEY) &&
535 nla_put_be32(skb, TCA_TUNNEL_KEY_ENC_KEY_ID, key_id)) ||
536 tunnel_key_dump_addresses(skb,
537 ¶ms->tcft_enc_metadata->u.tun_info) ||
539 nla_put_be16(skb, TCA_TUNNEL_KEY_ENC_DST_PORT,
541 nla_put_u8(skb, TCA_TUNNEL_KEY_NO_CSUM,
542 !(key->tun_flags & TUNNEL_CSUM)) ||
543 tunnel_key_opts_dump(skb, info))
544 goto nla_put_failure;
546 if (key->tos && nla_put_u8(skb, TCA_TUNNEL_KEY_ENC_TOS, key->tos))
547 goto nla_put_failure;
549 if (key->ttl && nla_put_u8(skb, TCA_TUNNEL_KEY_ENC_TTL, key->ttl))
550 goto nla_put_failure;
553 tcf_tm_dump(&tm, &t->tcf_tm);
554 if (nla_put_64bit(skb, TCA_TUNNEL_KEY_TM, sizeof(tm),
555 &tm, TCA_TUNNEL_KEY_PAD))
556 goto nla_put_failure;
557 spin_unlock_bh(&t->tcf_lock);
562 spin_unlock_bh(&t->tcf_lock);
567 static int tunnel_key_walker(struct net *net, struct sk_buff *skb,
568 struct netlink_callback *cb, int type,
569 const struct tc_action_ops *ops,
570 struct netlink_ext_ack *extack)
572 struct tc_action_net *tn = net_generic(net, tunnel_key_net_id);
574 return tcf_generic_walker(tn, skb, cb, type, ops, extack);
577 static int tunnel_key_search(struct net *net, struct tc_action **a, u32 index)
579 struct tc_action_net *tn = net_generic(net, tunnel_key_net_id);
581 return tcf_idr_search(tn, a, index);
584 static struct tc_action_ops act_tunnel_key_ops = {
585 .kind = "tunnel_key",
586 .id = TCA_ID_TUNNEL_KEY,
587 .owner = THIS_MODULE,
588 .act = tunnel_key_act,
589 .dump = tunnel_key_dump,
590 .init = tunnel_key_init,
591 .cleanup = tunnel_key_release,
592 .walk = tunnel_key_walker,
593 .lookup = tunnel_key_search,
594 .size = sizeof(struct tcf_tunnel_key),
597 static __net_init int tunnel_key_init_net(struct net *net)
599 struct tc_action_net *tn = net_generic(net, tunnel_key_net_id);
601 return tc_action_net_init(tn, &act_tunnel_key_ops);
604 static void __net_exit tunnel_key_exit_net(struct list_head *net_list)
606 tc_action_net_exit(net_list, tunnel_key_net_id);
609 static struct pernet_operations tunnel_key_net_ops = {
610 .init = tunnel_key_init_net,
611 .exit_batch = tunnel_key_exit_net,
612 .id = &tunnel_key_net_id,
613 .size = sizeof(struct tc_action_net),
616 static int __init tunnel_key_init_module(void)
618 return tcf_register_action(&act_tunnel_key_ops, &tunnel_key_net_ops);
621 static void __exit tunnel_key_cleanup_module(void)
623 tcf_unregister_action(&act_tunnel_key_ops, &tunnel_key_net_ops);
626 module_init(tunnel_key_init_module);
627 module_exit(tunnel_key_cleanup_module);
629 MODULE_AUTHOR("Amir Vadai <amir@vadai.me>");
630 MODULE_DESCRIPTION("ip tunnel manipulation actions");
631 MODULE_LICENSE("GPL v2");