2 * Copyright (c) 2016, Amir Vadai <amir@vadai.me>
3 * Copyright (c) 2016, Mellanox Technologies. All rights reserved.
5 * This program is free software; you can redistribute it and/or modify
6 * it under the terms of the GNU General Public License as published by
7 * the Free Software Foundation; either version 2 of the License, or
8 * (at your option) any later version.
11 #include <linux/module.h>
12 #include <linux/init.h>
13 #include <linux/kernel.h>
14 #include <linux/skbuff.h>
15 #include <linux/rtnetlink.h>
16 #include <net/geneve.h>
17 #include <net/netlink.h>
18 #include <net/pkt_sched.h>
21 #include <linux/tc_act/tc_tunnel_key.h>
22 #include <net/tc_act/tc_tunnel_key.h>
24 static unsigned int tunnel_key_net_id
;
25 static struct tc_action_ops act_tunnel_key_ops
;
27 static int tunnel_key_act(struct sk_buff
*skb
, const struct tc_action
*a
,
28 struct tcf_result
*res
)
30 struct tcf_tunnel_key
*t
= to_tunnel_key(a
);
31 struct tcf_tunnel_key_params
*params
;
34 params
= rcu_dereference_bh(t
->params
);
36 tcf_lastuse_update(&t
->tcf_tm
);
37 bstats_cpu_update(this_cpu_ptr(t
->common
.cpu_bstats
), skb
);
38 action
= READ_ONCE(t
->tcf_action
);
40 switch (params
->tcft_action
) {
41 case TCA_TUNNEL_KEY_ACT_RELEASE
:
44 case TCA_TUNNEL_KEY_ACT_SET
:
46 skb_dst_set(skb
, dst_clone(¶ms
->tcft_enc_metadata
->dst
));
49 WARN_ONCE(1, "Bad tunnel_key action %d.\n",
57 static const struct nla_policy
58 enc_opts_policy
[TCA_TUNNEL_KEY_ENC_OPTS_MAX
+ 1] = {
59 [TCA_TUNNEL_KEY_ENC_OPTS_GENEVE
] = { .type
= NLA_NESTED
},
62 static const struct nla_policy
63 geneve_opt_policy
[TCA_TUNNEL_KEY_ENC_OPT_GENEVE_MAX
+ 1] = {
64 [TCA_TUNNEL_KEY_ENC_OPT_GENEVE_CLASS
] = { .type
= NLA_U16
},
65 [TCA_TUNNEL_KEY_ENC_OPT_GENEVE_TYPE
] = { .type
= NLA_U8
},
66 [TCA_TUNNEL_KEY_ENC_OPT_GENEVE_DATA
] = { .type
= NLA_BINARY
,
71 tunnel_key_copy_geneve_opt(const struct nlattr
*nla
, void *dst
, int dst_len
,
72 struct netlink_ext_ack
*extack
)
74 struct nlattr
*tb
[TCA_TUNNEL_KEY_ENC_OPT_GENEVE_MAX
+ 1];
75 int err
, data_len
, opt_len
;
78 err
= nla_parse_nested(tb
, TCA_TUNNEL_KEY_ENC_OPT_GENEVE_MAX
,
79 nla
, geneve_opt_policy
, extack
);
83 if (!tb
[TCA_TUNNEL_KEY_ENC_OPT_GENEVE_CLASS
] ||
84 !tb
[TCA_TUNNEL_KEY_ENC_OPT_GENEVE_TYPE
] ||
85 !tb
[TCA_TUNNEL_KEY_ENC_OPT_GENEVE_DATA
]) {
86 NL_SET_ERR_MSG(extack
, "Missing tunnel key geneve option class, type or data");
90 data
= nla_data(tb
[TCA_TUNNEL_KEY_ENC_OPT_GENEVE_DATA
]);
91 data_len
= nla_len(tb
[TCA_TUNNEL_KEY_ENC_OPT_GENEVE_DATA
]);
93 NL_SET_ERR_MSG(extack
, "Tunnel key geneve option data is less than 4 bytes long");
97 NL_SET_ERR_MSG(extack
, "Tunnel key geneve option data is not a multiple of 4 bytes long");
101 opt_len
= sizeof(struct geneve_opt
) + data_len
;
103 struct geneve_opt
*opt
= dst
;
105 WARN_ON(dst_len
< opt_len
);
108 nla_get_be16(tb
[TCA_TUNNEL_KEY_ENC_OPT_GENEVE_CLASS
]);
109 opt
->type
= nla_get_u8(tb
[TCA_TUNNEL_KEY_ENC_OPT_GENEVE_TYPE
]);
110 opt
->length
= data_len
/ 4; /* length is in units of 4 bytes */
115 memcpy(opt
+ 1, data
, data_len
);
121 static int tunnel_key_copy_opts(const struct nlattr
*nla
, u8
*dst
,
122 int dst_len
, struct netlink_ext_ack
*extack
)
124 int err
, rem
, opt_len
, len
= nla_len(nla
), opts_len
= 0;
125 const struct nlattr
*attr
, *head
= nla_data(nla
);
127 err
= nla_validate(head
, len
, TCA_TUNNEL_KEY_ENC_OPTS_MAX
,
128 enc_opts_policy
, extack
);
132 nla_for_each_attr(attr
, head
, len
, rem
) {
133 switch (nla_type(attr
)) {
134 case TCA_TUNNEL_KEY_ENC_OPTS_GENEVE
:
135 opt_len
= tunnel_key_copy_geneve_opt(attr
, dst
,
149 NL_SET_ERR_MSG(extack
, "Empty list of tunnel options");
154 NL_SET_ERR_MSG(extack
, "Trailing data after parsing tunnel key options attributes");
161 static int tunnel_key_get_opts_len(struct nlattr
*nla
,
162 struct netlink_ext_ack
*extack
)
164 return tunnel_key_copy_opts(nla
, NULL
, 0, extack
);
167 static int tunnel_key_opts_set(struct nlattr
*nla
, struct ip_tunnel_info
*info
,
168 int opts_len
, struct netlink_ext_ack
*extack
)
170 info
->options_len
= opts_len
;
171 switch (nla_type(nla_data(nla
))) {
172 case TCA_TUNNEL_KEY_ENC_OPTS_GENEVE
:
173 #if IS_ENABLED(CONFIG_INET)
174 info
->key
.tun_flags
|= TUNNEL_GENEVE_OPT
;
175 return tunnel_key_copy_opts(nla
, ip_tunnel_info_opts(info
),
178 return -EAFNOSUPPORT
;
181 NL_SET_ERR_MSG(extack
, "Cannot set tunnel options for unknown tunnel type");
186 static const struct nla_policy tunnel_key_policy
[TCA_TUNNEL_KEY_MAX
+ 1] = {
187 [TCA_TUNNEL_KEY_PARMS
] = { .len
= sizeof(struct tc_tunnel_key
) },
188 [TCA_TUNNEL_KEY_ENC_IPV4_SRC
] = { .type
= NLA_U32
},
189 [TCA_TUNNEL_KEY_ENC_IPV4_DST
] = { .type
= NLA_U32
},
190 [TCA_TUNNEL_KEY_ENC_IPV6_SRC
] = { .len
= sizeof(struct in6_addr
) },
191 [TCA_TUNNEL_KEY_ENC_IPV6_DST
] = { .len
= sizeof(struct in6_addr
) },
192 [TCA_TUNNEL_KEY_ENC_KEY_ID
] = { .type
= NLA_U32
},
193 [TCA_TUNNEL_KEY_ENC_DST_PORT
] = {.type
= NLA_U16
},
194 [TCA_TUNNEL_KEY_NO_CSUM
] = { .type
= NLA_U8
},
195 [TCA_TUNNEL_KEY_ENC_OPTS
] = { .type
= NLA_NESTED
},
196 [TCA_TUNNEL_KEY_ENC_TOS
] = { .type
= NLA_U8
},
197 [TCA_TUNNEL_KEY_ENC_TTL
] = { .type
= NLA_U8
},
200 static int tunnel_key_init(struct net
*net
, struct nlattr
*nla
,
201 struct nlattr
*est
, struct tc_action
**a
,
202 int ovr
, int bind
, bool rtnl_held
,
203 struct netlink_ext_ack
*extack
)
205 struct tc_action_net
*tn
= net_generic(net
, tunnel_key_net_id
);
206 struct nlattr
*tb
[TCA_TUNNEL_KEY_MAX
+ 1];
207 struct tcf_tunnel_key_params
*params_new
;
208 struct metadata_dst
*metadata
= NULL
;
209 struct tc_tunnel_key
*parm
;
210 struct tcf_tunnel_key
*t
;
221 NL_SET_ERR_MSG(extack
, "Tunnel requires attributes to be passed");
225 err
= nla_parse_nested(tb
, TCA_TUNNEL_KEY_MAX
, nla
, tunnel_key_policy
,
228 NL_SET_ERR_MSG(extack
, "Failed to parse nested tunnel key attributes");
232 if (!tb
[TCA_TUNNEL_KEY_PARMS
]) {
233 NL_SET_ERR_MSG(extack
, "Missing tunnel key parameters");
237 parm
= nla_data(tb
[TCA_TUNNEL_KEY_PARMS
]);
238 err
= tcf_idr_check_alloc(tn
, &parm
->index
, a
, bind
);
245 switch (parm
->t_action
) {
246 case TCA_TUNNEL_KEY_ACT_RELEASE
:
248 case TCA_TUNNEL_KEY_ACT_SET
:
249 if (!tb
[TCA_TUNNEL_KEY_ENC_KEY_ID
]) {
250 NL_SET_ERR_MSG(extack
, "Missing tunnel key id");
255 key_id
= key32_to_tunnel_id(nla_get_be32(tb
[TCA_TUNNEL_KEY_ENC_KEY_ID
]));
257 flags
= TUNNEL_KEY
| TUNNEL_CSUM
;
258 if (tb
[TCA_TUNNEL_KEY_NO_CSUM
] &&
259 nla_get_u8(tb
[TCA_TUNNEL_KEY_NO_CSUM
]))
260 flags
&= ~TUNNEL_CSUM
;
262 if (tb
[TCA_TUNNEL_KEY_ENC_DST_PORT
])
263 dst_port
= nla_get_be16(tb
[TCA_TUNNEL_KEY_ENC_DST_PORT
]);
265 if (tb
[TCA_TUNNEL_KEY_ENC_OPTS
]) {
266 opts_len
= tunnel_key_get_opts_len(tb
[TCA_TUNNEL_KEY_ENC_OPTS
],
275 if (tb
[TCA_TUNNEL_KEY_ENC_TOS
])
276 tos
= nla_get_u8(tb
[TCA_TUNNEL_KEY_ENC_TOS
]);
278 if (tb
[TCA_TUNNEL_KEY_ENC_TTL
])
279 ttl
= nla_get_u8(tb
[TCA_TUNNEL_KEY_ENC_TTL
]);
281 if (tb
[TCA_TUNNEL_KEY_ENC_IPV4_SRC
] &&
282 tb
[TCA_TUNNEL_KEY_ENC_IPV4_DST
]) {
286 saddr
= nla_get_in_addr(tb
[TCA_TUNNEL_KEY_ENC_IPV4_SRC
]);
287 daddr
= nla_get_in_addr(tb
[TCA_TUNNEL_KEY_ENC_IPV4_DST
]);
289 metadata
= __ip_tun_set_dst(saddr
, daddr
, tos
, ttl
,
292 } else if (tb
[TCA_TUNNEL_KEY_ENC_IPV6_SRC
] &&
293 tb
[TCA_TUNNEL_KEY_ENC_IPV6_DST
]) {
294 struct in6_addr saddr
;
295 struct in6_addr daddr
;
297 saddr
= nla_get_in6_addr(tb
[TCA_TUNNEL_KEY_ENC_IPV6_SRC
]);
298 daddr
= nla_get_in6_addr(tb
[TCA_TUNNEL_KEY_ENC_IPV6_DST
]);
300 metadata
= __ipv6_tun_set_dst(&saddr
, &daddr
, tos
, ttl
, dst_port
,
304 NL_SET_ERR_MSG(extack
, "Missing either ipv4 or ipv6 src and dst");
310 NL_SET_ERR_MSG(extack
, "Cannot allocate tunnel metadata dst");
316 ret
= tunnel_key_opts_set(tb
[TCA_TUNNEL_KEY_ENC_OPTS
],
317 &metadata
->u
.tun_info
,
320 goto release_tun_meta
;
323 metadata
->u
.tun_info
.mode
|= IP_TUNNEL_INFO_TX
;
326 NL_SET_ERR_MSG(extack
, "Unknown tunnel key action");
332 ret
= tcf_idr_create(tn
, parm
->index
, est
, a
,
333 &act_tunnel_key_ops
, bind
, true);
335 NL_SET_ERR_MSG(extack
, "Cannot create TC IDR");
336 goto release_tun_meta
;
341 NL_SET_ERR_MSG(extack
, "TC IDR already exists");
343 goto release_tun_meta
;
346 t
= to_tunnel_key(*a
);
348 params_new
= kzalloc(sizeof(*params_new
), GFP_KERNEL
);
349 if (unlikely(!params_new
)) {
350 NL_SET_ERR_MSG(extack
, "Cannot allocate tunnel key parameters");
353 goto release_tun_meta
;
355 params_new
->tcft_action
= parm
->t_action
;
356 params_new
->tcft_enc_metadata
= metadata
;
358 spin_lock_bh(&t
->tcf_lock
);
359 t
->tcf_action
= parm
->action
;
360 rcu_swap_protected(t
->params
, params_new
,
361 lockdep_is_held(&t
->tcf_lock
));
362 spin_unlock_bh(&t
->tcf_lock
);
364 kfree_rcu(params_new
, rcu
);
366 if (ret
== ACT_P_CREATED
)
367 tcf_idr_insert(tn
, *a
);
372 dst_release(&metadata
->dst
);
376 tcf_idr_release(*a
, bind
);
378 tcf_idr_cleanup(tn
, parm
->index
);
382 static void tunnel_key_release(struct tc_action
*a
)
384 struct tcf_tunnel_key
*t
= to_tunnel_key(a
);
385 struct tcf_tunnel_key_params
*params
;
387 params
= rcu_dereference_protected(t
->params
, 1);
389 if (params
->tcft_action
== TCA_TUNNEL_KEY_ACT_SET
)
390 dst_release(¶ms
->tcft_enc_metadata
->dst
);
392 kfree_rcu(params
, rcu
);
396 static int tunnel_key_geneve_opts_dump(struct sk_buff
*skb
,
397 const struct ip_tunnel_info
*info
)
399 int len
= info
->options_len
;
400 u8
*src
= (u8
*)(info
+ 1);
401 struct nlattr
*start
;
403 start
= nla_nest_start(skb
, TCA_TUNNEL_KEY_ENC_OPTS_GENEVE
);
408 struct geneve_opt
*opt
= (struct geneve_opt
*)src
;
410 if (nla_put_be16(skb
, TCA_TUNNEL_KEY_ENC_OPT_GENEVE_CLASS
,
412 nla_put_u8(skb
, TCA_TUNNEL_KEY_ENC_OPT_GENEVE_TYPE
,
414 nla_put(skb
, TCA_TUNNEL_KEY_ENC_OPT_GENEVE_DATA
,
415 opt
->length
* 4, opt
+ 1)) {
416 nla_nest_cancel(skb
, start
);
420 len
-= sizeof(struct geneve_opt
) + opt
->length
* 4;
421 src
+= sizeof(struct geneve_opt
) + opt
->length
* 4;
424 nla_nest_end(skb
, start
);
428 static int tunnel_key_opts_dump(struct sk_buff
*skb
,
429 const struct ip_tunnel_info
*info
)
431 struct nlattr
*start
;
434 if (!info
->options_len
)
437 start
= nla_nest_start(skb
, TCA_TUNNEL_KEY_ENC_OPTS
);
441 if (info
->key
.tun_flags
& TUNNEL_GENEVE_OPT
) {
442 err
= tunnel_key_geneve_opts_dump(skb
, info
);
447 nla_nest_cancel(skb
, start
);
451 nla_nest_end(skb
, start
);
455 static int tunnel_key_dump_addresses(struct sk_buff
*skb
,
456 const struct ip_tunnel_info
*info
)
458 unsigned short family
= ip_tunnel_info_af(info
);
460 if (family
== AF_INET
) {
461 __be32 saddr
= info
->key
.u
.ipv4
.src
;
462 __be32 daddr
= info
->key
.u
.ipv4
.dst
;
464 if (!nla_put_in_addr(skb
, TCA_TUNNEL_KEY_ENC_IPV4_SRC
, saddr
) &&
465 !nla_put_in_addr(skb
, TCA_TUNNEL_KEY_ENC_IPV4_DST
, daddr
))
469 if (family
== AF_INET6
) {
470 const struct in6_addr
*saddr6
= &info
->key
.u
.ipv6
.src
;
471 const struct in6_addr
*daddr6
= &info
->key
.u
.ipv6
.dst
;
473 if (!nla_put_in6_addr(skb
,
474 TCA_TUNNEL_KEY_ENC_IPV6_SRC
, saddr6
) &&
475 !nla_put_in6_addr(skb
,
476 TCA_TUNNEL_KEY_ENC_IPV6_DST
, daddr6
))
483 static int tunnel_key_dump(struct sk_buff
*skb
, struct tc_action
*a
,
486 unsigned char *b
= skb_tail_pointer(skb
);
487 struct tcf_tunnel_key
*t
= to_tunnel_key(a
);
488 struct tcf_tunnel_key_params
*params
;
489 struct tc_tunnel_key opt
= {
490 .index
= t
->tcf_index
,
491 .refcnt
= refcount_read(&t
->tcf_refcnt
) - ref
,
492 .bindcnt
= atomic_read(&t
->tcf_bindcnt
) - bind
,
496 spin_lock_bh(&t
->tcf_lock
);
497 params
= rcu_dereference_protected(t
->params
,
498 lockdep_is_held(&t
->tcf_lock
));
499 opt
.action
= t
->tcf_action
;
500 opt
.t_action
= params
->tcft_action
;
502 if (nla_put(skb
, TCA_TUNNEL_KEY_PARMS
, sizeof(opt
), &opt
))
503 goto nla_put_failure
;
505 if (params
->tcft_action
== TCA_TUNNEL_KEY_ACT_SET
) {
506 struct ip_tunnel_info
*info
=
507 ¶ms
->tcft_enc_metadata
->u
.tun_info
;
508 struct ip_tunnel_key
*key
= &info
->key
;
509 __be32 key_id
= tunnel_id_to_key32(key
->tun_id
);
511 if (nla_put_be32(skb
, TCA_TUNNEL_KEY_ENC_KEY_ID
, key_id
) ||
512 tunnel_key_dump_addresses(skb
,
513 ¶ms
->tcft_enc_metadata
->u
.tun_info
) ||
514 nla_put_be16(skb
, TCA_TUNNEL_KEY_ENC_DST_PORT
, key
->tp_dst
) ||
515 nla_put_u8(skb
, TCA_TUNNEL_KEY_NO_CSUM
,
516 !(key
->tun_flags
& TUNNEL_CSUM
)) ||
517 tunnel_key_opts_dump(skb
, info
))
518 goto nla_put_failure
;
520 if (key
->tos
&& nla_put_u8(skb
, TCA_TUNNEL_KEY_ENC_TOS
, key
->tos
))
521 goto nla_put_failure
;
523 if (key
->ttl
&& nla_put_u8(skb
, TCA_TUNNEL_KEY_ENC_TTL
, key
->ttl
))
524 goto nla_put_failure
;
527 tcf_tm_dump(&tm
, &t
->tcf_tm
);
528 if (nla_put_64bit(skb
, TCA_TUNNEL_KEY_TM
, sizeof(tm
),
529 &tm
, TCA_TUNNEL_KEY_PAD
))
530 goto nla_put_failure
;
531 spin_unlock_bh(&t
->tcf_lock
);
536 spin_unlock_bh(&t
->tcf_lock
);
541 static int tunnel_key_walker(struct net
*net
, struct sk_buff
*skb
,
542 struct netlink_callback
*cb
, int type
,
543 const struct tc_action_ops
*ops
,
544 struct netlink_ext_ack
*extack
)
546 struct tc_action_net
*tn
= net_generic(net
, tunnel_key_net_id
);
548 return tcf_generic_walker(tn
, skb
, cb
, type
, ops
, extack
);
551 static int tunnel_key_search(struct net
*net
, struct tc_action
**a
, u32 index
)
553 struct tc_action_net
*tn
= net_generic(net
, tunnel_key_net_id
);
555 return tcf_idr_search(tn
, a
, index
);
558 static struct tc_action_ops act_tunnel_key_ops
= {
559 .kind
= "tunnel_key",
560 .type
= TCA_ACT_TUNNEL_KEY
,
561 .owner
= THIS_MODULE
,
562 .act
= tunnel_key_act
,
563 .dump
= tunnel_key_dump
,
564 .init
= tunnel_key_init
,
565 .cleanup
= tunnel_key_release
,
566 .walk
= tunnel_key_walker
,
567 .lookup
= tunnel_key_search
,
568 .size
= sizeof(struct tcf_tunnel_key
),
571 static __net_init
int tunnel_key_init_net(struct net
*net
)
573 struct tc_action_net
*tn
= net_generic(net
, tunnel_key_net_id
);
575 return tc_action_net_init(tn
, &act_tunnel_key_ops
);
578 static void __net_exit
tunnel_key_exit_net(struct list_head
*net_list
)
580 tc_action_net_exit(net_list
, tunnel_key_net_id
);
583 static struct pernet_operations tunnel_key_net_ops
= {
584 .init
= tunnel_key_init_net
,
585 .exit_batch
= tunnel_key_exit_net
,
586 .id
= &tunnel_key_net_id
,
587 .size
= sizeof(struct tc_action_net
),
590 static int __init
tunnel_key_init_module(void)
592 return tcf_register_action(&act_tunnel_key_ops
, &tunnel_key_net_ops
);
595 static void __exit
tunnel_key_cleanup_module(void)
597 tcf_unregister_action(&act_tunnel_key_ops
, &tunnel_key_net_ops
);
600 module_init(tunnel_key_init_module
);
601 module_exit(tunnel_key_cleanup_module
);
603 MODULE_AUTHOR("Amir Vadai <amir@vadai.me>");
604 MODULE_DESCRIPTION("ip tunnel manipulation actions");
605 MODULE_LICENSE("GPL v2");