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
,
140 if (opts_len
> IP_TUNNEL_OPTS_MAX
) {
141 NL_SET_ERR_MSG(extack
, "Tunnel options exceeds max size");
153 NL_SET_ERR_MSG(extack
, "Empty list of tunnel options");
158 NL_SET_ERR_MSG(extack
, "Trailing data after parsing tunnel key options attributes");
165 static int tunnel_key_get_opts_len(struct nlattr
*nla
,
166 struct netlink_ext_ack
*extack
)
168 return tunnel_key_copy_opts(nla
, NULL
, 0, extack
);
171 static int tunnel_key_opts_set(struct nlattr
*nla
, struct ip_tunnel_info
*info
,
172 int opts_len
, struct netlink_ext_ack
*extack
)
174 info
->options_len
= opts_len
;
175 switch (nla_type(nla_data(nla
))) {
176 case TCA_TUNNEL_KEY_ENC_OPTS_GENEVE
:
177 #if IS_ENABLED(CONFIG_INET)
178 info
->key
.tun_flags
|= TUNNEL_GENEVE_OPT
;
179 return tunnel_key_copy_opts(nla
, ip_tunnel_info_opts(info
),
182 return -EAFNOSUPPORT
;
185 NL_SET_ERR_MSG(extack
, "Cannot set tunnel options for unknown tunnel type");
190 static const struct nla_policy tunnel_key_policy
[TCA_TUNNEL_KEY_MAX
+ 1] = {
191 [TCA_TUNNEL_KEY_PARMS
] = { .len
= sizeof(struct tc_tunnel_key
) },
192 [TCA_TUNNEL_KEY_ENC_IPV4_SRC
] = { .type
= NLA_U32
},
193 [TCA_TUNNEL_KEY_ENC_IPV4_DST
] = { .type
= NLA_U32
},
194 [TCA_TUNNEL_KEY_ENC_IPV6_SRC
] = { .len
= sizeof(struct in6_addr
) },
195 [TCA_TUNNEL_KEY_ENC_IPV6_DST
] = { .len
= sizeof(struct in6_addr
) },
196 [TCA_TUNNEL_KEY_ENC_KEY_ID
] = { .type
= NLA_U32
},
197 [TCA_TUNNEL_KEY_ENC_DST_PORT
] = {.type
= NLA_U16
},
198 [TCA_TUNNEL_KEY_NO_CSUM
] = { .type
= NLA_U8
},
199 [TCA_TUNNEL_KEY_ENC_OPTS
] = { .type
= NLA_NESTED
},
200 [TCA_TUNNEL_KEY_ENC_TOS
] = { .type
= NLA_U8
},
201 [TCA_TUNNEL_KEY_ENC_TTL
] = { .type
= NLA_U8
},
204 static void tunnel_key_release_params(struct tcf_tunnel_key_params
*p
)
208 if (p
->tcft_action
== TCA_TUNNEL_KEY_ACT_SET
)
209 dst_release(&p
->tcft_enc_metadata
->dst
);
213 static int tunnel_key_init(struct net
*net
, struct nlattr
*nla
,
214 struct nlattr
*est
, struct tc_action
**a
,
215 int ovr
, int bind
, bool rtnl_held
,
216 struct netlink_ext_ack
*extack
)
218 struct tc_action_net
*tn
= net_generic(net
, tunnel_key_net_id
);
219 struct nlattr
*tb
[TCA_TUNNEL_KEY_MAX
+ 1];
220 struct tcf_tunnel_key_params
*params_new
;
221 struct metadata_dst
*metadata
= NULL
;
222 struct tc_tunnel_key
*parm
;
223 struct tcf_tunnel_key
*t
;
235 NL_SET_ERR_MSG(extack
, "Tunnel requires attributes to be passed");
239 err
= nla_parse_nested(tb
, TCA_TUNNEL_KEY_MAX
, nla
, tunnel_key_policy
,
242 NL_SET_ERR_MSG(extack
, "Failed to parse nested tunnel key attributes");
246 if (!tb
[TCA_TUNNEL_KEY_PARMS
]) {
247 NL_SET_ERR_MSG(extack
, "Missing tunnel key parameters");
251 parm
= nla_data(tb
[TCA_TUNNEL_KEY_PARMS
]);
253 err
= tcf_idr_check_alloc(tn
, &index
, a
, bind
);
260 switch (parm
->t_action
) {
261 case TCA_TUNNEL_KEY_ACT_RELEASE
:
263 case TCA_TUNNEL_KEY_ACT_SET
:
264 if (!tb
[TCA_TUNNEL_KEY_ENC_KEY_ID
]) {
265 NL_SET_ERR_MSG(extack
, "Missing tunnel key id");
270 key_id
= key32_to_tunnel_id(nla_get_be32(tb
[TCA_TUNNEL_KEY_ENC_KEY_ID
]));
272 flags
= TUNNEL_KEY
| TUNNEL_CSUM
;
273 if (tb
[TCA_TUNNEL_KEY_NO_CSUM
] &&
274 nla_get_u8(tb
[TCA_TUNNEL_KEY_NO_CSUM
]))
275 flags
&= ~TUNNEL_CSUM
;
277 if (tb
[TCA_TUNNEL_KEY_ENC_DST_PORT
])
278 dst_port
= nla_get_be16(tb
[TCA_TUNNEL_KEY_ENC_DST_PORT
]);
280 if (tb
[TCA_TUNNEL_KEY_ENC_OPTS
]) {
281 opts_len
= tunnel_key_get_opts_len(tb
[TCA_TUNNEL_KEY_ENC_OPTS
],
290 if (tb
[TCA_TUNNEL_KEY_ENC_TOS
])
291 tos
= nla_get_u8(tb
[TCA_TUNNEL_KEY_ENC_TOS
]);
293 if (tb
[TCA_TUNNEL_KEY_ENC_TTL
])
294 ttl
= nla_get_u8(tb
[TCA_TUNNEL_KEY_ENC_TTL
]);
296 if (tb
[TCA_TUNNEL_KEY_ENC_IPV4_SRC
] &&
297 tb
[TCA_TUNNEL_KEY_ENC_IPV4_DST
]) {
301 saddr
= nla_get_in_addr(tb
[TCA_TUNNEL_KEY_ENC_IPV4_SRC
]);
302 daddr
= nla_get_in_addr(tb
[TCA_TUNNEL_KEY_ENC_IPV4_DST
]);
304 metadata
= __ip_tun_set_dst(saddr
, daddr
, tos
, ttl
,
307 } else if (tb
[TCA_TUNNEL_KEY_ENC_IPV6_SRC
] &&
308 tb
[TCA_TUNNEL_KEY_ENC_IPV6_DST
]) {
309 struct in6_addr saddr
;
310 struct in6_addr daddr
;
312 saddr
= nla_get_in6_addr(tb
[TCA_TUNNEL_KEY_ENC_IPV6_SRC
]);
313 daddr
= nla_get_in6_addr(tb
[TCA_TUNNEL_KEY_ENC_IPV6_DST
]);
315 metadata
= __ipv6_tun_set_dst(&saddr
, &daddr
, tos
, ttl
, dst_port
,
319 NL_SET_ERR_MSG(extack
, "Missing either ipv4 or ipv6 src and dst");
325 NL_SET_ERR_MSG(extack
, "Cannot allocate tunnel metadata dst");
331 ret
= tunnel_key_opts_set(tb
[TCA_TUNNEL_KEY_ENC_OPTS
],
332 &metadata
->u
.tun_info
,
335 goto release_tun_meta
;
338 metadata
->u
.tun_info
.mode
|= IP_TUNNEL_INFO_TX
;
341 NL_SET_ERR_MSG(extack
, "Unknown tunnel key action");
347 ret
= tcf_idr_create(tn
, index
, est
, a
,
348 &act_tunnel_key_ops
, bind
, true);
350 NL_SET_ERR_MSG(extack
, "Cannot create TC IDR");
351 goto release_tun_meta
;
356 NL_SET_ERR_MSG(extack
, "TC IDR already exists");
358 goto release_tun_meta
;
361 t
= to_tunnel_key(*a
);
363 params_new
= kzalloc(sizeof(*params_new
), GFP_KERNEL
);
364 if (unlikely(!params_new
)) {
365 NL_SET_ERR_MSG(extack
, "Cannot allocate tunnel key parameters");
368 goto release_tun_meta
;
370 params_new
->tcft_action
= parm
->t_action
;
371 params_new
->tcft_enc_metadata
= metadata
;
373 spin_lock_bh(&t
->tcf_lock
);
374 t
->tcf_action
= parm
->action
;
375 rcu_swap_protected(t
->params
, params_new
,
376 lockdep_is_held(&t
->tcf_lock
));
377 spin_unlock_bh(&t
->tcf_lock
);
378 tunnel_key_release_params(params_new
);
380 if (ret
== ACT_P_CREATED
)
381 tcf_idr_insert(tn
, *a
);
387 dst_release(&metadata
->dst
);
391 tcf_idr_release(*a
, bind
);
393 tcf_idr_cleanup(tn
, index
);
397 static void tunnel_key_release(struct tc_action
*a
)
399 struct tcf_tunnel_key
*t
= to_tunnel_key(a
);
400 struct tcf_tunnel_key_params
*params
;
402 params
= rcu_dereference_protected(t
->params
, 1);
403 tunnel_key_release_params(params
);
406 static int tunnel_key_geneve_opts_dump(struct sk_buff
*skb
,
407 const struct ip_tunnel_info
*info
)
409 int len
= info
->options_len
;
410 u8
*src
= (u8
*)(info
+ 1);
411 struct nlattr
*start
;
413 start
= nla_nest_start(skb
, TCA_TUNNEL_KEY_ENC_OPTS_GENEVE
);
418 struct geneve_opt
*opt
= (struct geneve_opt
*)src
;
420 if (nla_put_be16(skb
, TCA_TUNNEL_KEY_ENC_OPT_GENEVE_CLASS
,
422 nla_put_u8(skb
, TCA_TUNNEL_KEY_ENC_OPT_GENEVE_TYPE
,
424 nla_put(skb
, TCA_TUNNEL_KEY_ENC_OPT_GENEVE_DATA
,
425 opt
->length
* 4, opt
+ 1)) {
426 nla_nest_cancel(skb
, start
);
430 len
-= sizeof(struct geneve_opt
) + opt
->length
* 4;
431 src
+= sizeof(struct geneve_opt
) + opt
->length
* 4;
434 nla_nest_end(skb
, start
);
438 static int tunnel_key_opts_dump(struct sk_buff
*skb
,
439 const struct ip_tunnel_info
*info
)
441 struct nlattr
*start
;
444 if (!info
->options_len
)
447 start
= nla_nest_start(skb
, TCA_TUNNEL_KEY_ENC_OPTS
);
451 if (info
->key
.tun_flags
& TUNNEL_GENEVE_OPT
) {
452 err
= tunnel_key_geneve_opts_dump(skb
, info
);
457 nla_nest_cancel(skb
, start
);
461 nla_nest_end(skb
, start
);
465 static int tunnel_key_dump_addresses(struct sk_buff
*skb
,
466 const struct ip_tunnel_info
*info
)
468 unsigned short family
= ip_tunnel_info_af(info
);
470 if (family
== AF_INET
) {
471 __be32 saddr
= info
->key
.u
.ipv4
.src
;
472 __be32 daddr
= info
->key
.u
.ipv4
.dst
;
474 if (!nla_put_in_addr(skb
, TCA_TUNNEL_KEY_ENC_IPV4_SRC
, saddr
) &&
475 !nla_put_in_addr(skb
, TCA_TUNNEL_KEY_ENC_IPV4_DST
, daddr
))
479 if (family
== AF_INET6
) {
480 const struct in6_addr
*saddr6
= &info
->key
.u
.ipv6
.src
;
481 const struct in6_addr
*daddr6
= &info
->key
.u
.ipv6
.dst
;
483 if (!nla_put_in6_addr(skb
,
484 TCA_TUNNEL_KEY_ENC_IPV6_SRC
, saddr6
) &&
485 !nla_put_in6_addr(skb
,
486 TCA_TUNNEL_KEY_ENC_IPV6_DST
, daddr6
))
493 static int tunnel_key_dump(struct sk_buff
*skb
, struct tc_action
*a
,
496 unsigned char *b
= skb_tail_pointer(skb
);
497 struct tcf_tunnel_key
*t
= to_tunnel_key(a
);
498 struct tcf_tunnel_key_params
*params
;
499 struct tc_tunnel_key opt
= {
500 .index
= t
->tcf_index
,
501 .refcnt
= refcount_read(&t
->tcf_refcnt
) - ref
,
502 .bindcnt
= atomic_read(&t
->tcf_bindcnt
) - bind
,
506 spin_lock_bh(&t
->tcf_lock
);
507 params
= rcu_dereference_protected(t
->params
,
508 lockdep_is_held(&t
->tcf_lock
));
509 opt
.action
= t
->tcf_action
;
510 opt
.t_action
= params
->tcft_action
;
512 if (nla_put(skb
, TCA_TUNNEL_KEY_PARMS
, sizeof(opt
), &opt
))
513 goto nla_put_failure
;
515 if (params
->tcft_action
== TCA_TUNNEL_KEY_ACT_SET
) {
516 struct ip_tunnel_info
*info
=
517 ¶ms
->tcft_enc_metadata
->u
.tun_info
;
518 struct ip_tunnel_key
*key
= &info
->key
;
519 __be32 key_id
= tunnel_id_to_key32(key
->tun_id
);
521 if (nla_put_be32(skb
, TCA_TUNNEL_KEY_ENC_KEY_ID
, key_id
) ||
522 tunnel_key_dump_addresses(skb
,
523 ¶ms
->tcft_enc_metadata
->u
.tun_info
) ||
524 nla_put_be16(skb
, TCA_TUNNEL_KEY_ENC_DST_PORT
, key
->tp_dst
) ||
525 nla_put_u8(skb
, TCA_TUNNEL_KEY_NO_CSUM
,
526 !(key
->tun_flags
& TUNNEL_CSUM
)) ||
527 tunnel_key_opts_dump(skb
, info
))
528 goto nla_put_failure
;
530 if (key
->tos
&& nla_put_u8(skb
, TCA_TUNNEL_KEY_ENC_TOS
, key
->tos
))
531 goto nla_put_failure
;
533 if (key
->ttl
&& nla_put_u8(skb
, TCA_TUNNEL_KEY_ENC_TTL
, key
->ttl
))
534 goto nla_put_failure
;
537 tcf_tm_dump(&tm
, &t
->tcf_tm
);
538 if (nla_put_64bit(skb
, TCA_TUNNEL_KEY_TM
, sizeof(tm
),
539 &tm
, TCA_TUNNEL_KEY_PAD
))
540 goto nla_put_failure
;
541 spin_unlock_bh(&t
->tcf_lock
);
546 spin_unlock_bh(&t
->tcf_lock
);
551 static int tunnel_key_walker(struct net
*net
, struct sk_buff
*skb
,
552 struct netlink_callback
*cb
, int type
,
553 const struct tc_action_ops
*ops
,
554 struct netlink_ext_ack
*extack
)
556 struct tc_action_net
*tn
= net_generic(net
, tunnel_key_net_id
);
558 return tcf_generic_walker(tn
, skb
, cb
, type
, ops
, extack
);
561 static int tunnel_key_search(struct net
*net
, struct tc_action
**a
, u32 index
,
562 struct netlink_ext_ack
*extack
)
564 struct tc_action_net
*tn
= net_generic(net
, tunnel_key_net_id
);
566 return tcf_idr_search(tn
, a
, index
);
569 static struct tc_action_ops act_tunnel_key_ops
= {
570 .kind
= "tunnel_key",
571 .type
= TCA_ACT_TUNNEL_KEY
,
572 .owner
= THIS_MODULE
,
573 .act
= tunnel_key_act
,
574 .dump
= tunnel_key_dump
,
575 .init
= tunnel_key_init
,
576 .cleanup
= tunnel_key_release
,
577 .walk
= tunnel_key_walker
,
578 .lookup
= tunnel_key_search
,
579 .size
= sizeof(struct tcf_tunnel_key
),
582 static __net_init
int tunnel_key_init_net(struct net
*net
)
584 struct tc_action_net
*tn
= net_generic(net
, tunnel_key_net_id
);
586 return tc_action_net_init(net
, tn
, &act_tunnel_key_ops
);
589 static void __net_exit
tunnel_key_exit_net(struct list_head
*net_list
)
591 tc_action_net_exit(net_list
, tunnel_key_net_id
);
594 static struct pernet_operations tunnel_key_net_ops
= {
595 .init
= tunnel_key_init_net
,
596 .exit_batch
= tunnel_key_exit_net
,
597 .id
= &tunnel_key_net_id
,
598 .size
= sizeof(struct tc_action_net
),
601 static int __init
tunnel_key_init_module(void)
603 return tcf_register_action(&act_tunnel_key_ops
, &tunnel_key_net_ops
);
606 static void __exit
tunnel_key_cleanup_module(void)
608 tcf_unregister_action(&act_tunnel_key_ops
, &tunnel_key_net_ops
);
611 module_init(tunnel_key_init_module
);
612 module_exit(tunnel_key_cleanup_module
);
614 MODULE_AUTHOR("Amir Vadai <amir@vadai.me>");
615 MODULE_DESCRIPTION("ip tunnel manipulation actions");
616 MODULE_LICENSE("GPL v2");