1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef __NET_PKT_CLS_H
3 #define __NET_PKT_CLS_H
5 #include <linux/pkt_cls.h>
6 #include <linux/workqueue.h>
7 #include <net/sch_generic.h>
8 #include <net/act_api.h>
9 #include <net/net_namespace.h>
11 /* TC action not accessible from user space */
12 #define TC_ACT_CONSUMED (TC_ACT_VALUE_MAX + 1)
14 /* Basic packet classifier frontend definitions. */
22 int (*fn
)(struct tcf_proto
*, void *node
, struct tcf_walker
*);
25 int register_tcf_proto_ops(struct tcf_proto_ops
*ops
);
26 int unregister_tcf_proto_ops(struct tcf_proto_ops
*ops
);
28 struct tcf_block_ext_info
{
29 enum flow_block_binder_type binder_type
;
30 tcf_chain_head_change_t
*chain_head_change
;
31 void *chain_head_change_priv
;
36 bool tcf_queue_work(struct rcu_work
*rwork
, work_func_t func
);
39 struct tcf_chain
*tcf_chain_get_by_act(struct tcf_block
*block
,
41 void tcf_chain_put_by_act(struct tcf_chain
*chain
);
42 struct tcf_chain
*tcf_get_next_chain(struct tcf_block
*block
,
43 struct tcf_chain
*chain
);
44 struct tcf_proto
*tcf_get_next_proto(struct tcf_chain
*chain
,
45 struct tcf_proto
*tp
, bool rtnl_held
);
46 void tcf_block_netif_keep_dst(struct tcf_block
*block
);
47 int tcf_block_get(struct tcf_block
**p_block
,
48 struct tcf_proto __rcu
**p_filter_chain
, struct Qdisc
*q
,
49 struct netlink_ext_ack
*extack
);
50 int tcf_block_get_ext(struct tcf_block
**p_block
, struct Qdisc
*q
,
51 struct tcf_block_ext_info
*ei
,
52 struct netlink_ext_ack
*extack
);
53 void tcf_block_put(struct tcf_block
*block
);
54 void tcf_block_put_ext(struct tcf_block
*block
, struct Qdisc
*q
,
55 struct tcf_block_ext_info
*ei
);
57 static inline bool tcf_block_shared(struct tcf_block
*block
)
62 static inline bool tcf_block_non_null_shared(struct tcf_block
*block
)
64 return block
&& block
->index
;
67 static inline struct Qdisc
*tcf_block_q(struct tcf_block
*block
)
69 WARN_ON(tcf_block_shared(block
));
73 int tcf_classify(struct sk_buff
*skb
, const struct tcf_proto
*tp
,
74 struct tcf_result
*res
, bool compat_mode
);
75 int tcf_classify_ingress(struct sk_buff
*skb
,
76 const struct tcf_block
*ingress_block
,
77 const struct tcf_proto
*tp
, struct tcf_result
*res
,
81 static inline bool tcf_block_shared(struct tcf_block
*block
)
86 static inline bool tcf_block_non_null_shared(struct tcf_block
*block
)
92 int tcf_block_get(struct tcf_block
**p_block
,
93 struct tcf_proto __rcu
**p_filter_chain
, struct Qdisc
*q
,
94 struct netlink_ext_ack
*extack
)
100 int tcf_block_get_ext(struct tcf_block
**p_block
, struct Qdisc
*q
,
101 struct tcf_block_ext_info
*ei
,
102 struct netlink_ext_ack
*extack
)
107 static inline void tcf_block_put(struct tcf_block
*block
)
112 void tcf_block_put_ext(struct tcf_block
*block
, struct Qdisc
*q
,
113 struct tcf_block_ext_info
*ei
)
117 static inline struct Qdisc
*tcf_block_q(struct tcf_block
*block
)
123 int tc_setup_cb_block_register(struct tcf_block
*block
, flow_setup_cb_t
*cb
,
130 void tc_setup_cb_block_unregister(struct tcf_block
*block
, flow_setup_cb_t
*cb
,
135 static inline int tcf_classify(struct sk_buff
*skb
, const struct tcf_proto
*tp
,
136 struct tcf_result
*res
, bool compat_mode
)
138 return TC_ACT_UNSPEC
;
141 static inline int tcf_classify_ingress(struct sk_buff
*skb
,
142 const struct tcf_block
*ingress_block
,
143 const struct tcf_proto
*tp
,
144 struct tcf_result
*res
, bool compat_mode
)
146 return TC_ACT_UNSPEC
;
151 static inline unsigned long
152 __cls_set_class(unsigned long *clp
, unsigned long cl
)
154 return xchg(clp
, cl
);
158 __tcf_bind_filter(struct Qdisc
*q
, struct tcf_result
*r
, unsigned long base
)
162 cl
= q
->ops
->cl_ops
->bind_tcf(q
, base
, r
->classid
);
163 cl
= __cls_set_class(&r
->class, cl
);
165 q
->ops
->cl_ops
->unbind_tcf(q
, cl
);
169 tcf_bind_filter(struct tcf_proto
*tp
, struct tcf_result
*r
, unsigned long base
)
171 struct Qdisc
*q
= tp
->chain
->block
->q
;
173 /* Check q as it is not set for shared blocks. In that case,
174 * setting class is not supported.
179 __tcf_bind_filter(q
, r
, base
);
184 __tcf_unbind_filter(struct Qdisc
*q
, struct tcf_result
*r
)
188 if ((cl
= __cls_set_class(&r
->class, 0)) != 0)
189 q
->ops
->cl_ops
->unbind_tcf(q
, cl
);
193 tcf_unbind_filter(struct tcf_proto
*tp
, struct tcf_result
*r
)
195 struct Qdisc
*q
= tp
->chain
->block
->q
;
199 __tcf_unbind_filter(q
, r
);
203 #ifdef CONFIG_NET_CLS_ACT
204 __u32 type
; /* for backward compat(TCA_OLD_COMPAT) */
206 struct tc_action
**actions
;
209 /* Map to export classifier specific extension TLV types to the
210 * generic extensions API. Unsupported extensions must be set to 0.
216 static inline int tcf_exts_init(struct tcf_exts
*exts
, struct net
*net
,
217 int action
, int police
)
219 #ifdef CONFIG_NET_CLS_ACT
221 exts
->nr_actions
= 0;
223 exts
->actions
= kcalloc(TCA_ACT_MAX_PRIO
, sizeof(struct tc_action
*),
228 exts
->action
= action
;
229 exts
->police
= police
;
233 /* Return false if the netns is being destroyed in cleanup_net(). Callers
234 * need to do cleanup synchronously in this case, otherwise may race with
235 * tc_action_net_exit(). Return true for other cases.
237 static inline bool tcf_exts_get_net(struct tcf_exts
*exts
)
239 #ifdef CONFIG_NET_CLS_ACT
240 exts
->net
= maybe_get_net(exts
->net
);
241 return exts
->net
!= NULL
;
247 static inline void tcf_exts_put_net(struct tcf_exts
*exts
)
249 #ifdef CONFIG_NET_CLS_ACT
255 #ifdef CONFIG_NET_CLS_ACT
256 #define tcf_exts_for_each_action(i, a, exts) \
257 for (i = 0; i < TCA_ACT_MAX_PRIO && ((a) = (exts)->actions[i]); i++)
259 #define tcf_exts_for_each_action(i, a, exts) \
260 for (; 0; (void)(i), (void)(a), (void)(exts))
264 tcf_exts_stats_update(const struct tcf_exts
*exts
,
265 u64 bytes
, u64 packets
, u64 lastuse
,
266 u8 used_hw_stats
, bool used_hw_stats_valid
)
268 #ifdef CONFIG_NET_CLS_ACT
273 for (i
= 0; i
< exts
->nr_actions
; i
++) {
274 struct tc_action
*a
= exts
->actions
[i
];
276 tcf_action_stats_update(a
, bytes
, packets
, lastuse
, true);
277 a
->used_hw_stats
= used_hw_stats
;
278 a
->used_hw_stats_valid
= used_hw_stats_valid
;
286 * tcf_exts_has_actions - check if at least one action is present
287 * @exts: tc filter extensions handle
289 * Returns true if at least one action is present.
291 static inline bool tcf_exts_has_actions(struct tcf_exts
*exts
)
293 #ifdef CONFIG_NET_CLS_ACT
294 return exts
->nr_actions
;
301 * tcf_exts_exec - execute tc filter extensions
302 * @skb: socket buffer
303 * @exts: tc filter extensions handle
304 * @res: desired result
306 * Executes all configured extensions. Returns TC_ACT_OK on a normal execution,
307 * a negative number if the filter must be considered unmatched or
308 * a positive action code (TC_ACT_*) which must be returned to the
312 tcf_exts_exec(struct sk_buff
*skb
, struct tcf_exts
*exts
,
313 struct tcf_result
*res
)
315 #ifdef CONFIG_NET_CLS_ACT
316 return tcf_action_exec(skb
, exts
->actions
, exts
->nr_actions
, res
);
321 int tcf_exts_validate(struct net
*net
, struct tcf_proto
*tp
,
322 struct nlattr
**tb
, struct nlattr
*rate_tlv
,
323 struct tcf_exts
*exts
, bool ovr
, bool rtnl_held
,
324 struct netlink_ext_ack
*extack
);
325 void tcf_exts_destroy(struct tcf_exts
*exts
);
326 void tcf_exts_change(struct tcf_exts
*dst
, struct tcf_exts
*src
);
327 int tcf_exts_dump(struct sk_buff
*skb
, struct tcf_exts
*exts
);
328 int tcf_exts_dump_stats(struct sk_buff
*skb
, struct tcf_exts
*exts
);
331 * struct tcf_pkt_info - packet information
333 struct tcf_pkt_info
{
338 #ifdef CONFIG_NET_EMATCH
340 struct tcf_ematch_ops
;
343 * struct tcf_ematch - extended match (ematch)
345 * @matchid: identifier to allow userspace to reidentify a match
346 * @flags: flags specifying attributes and the relation to other matches
347 * @ops: the operations lookup table of the corresponding ematch module
348 * @datalen: length of the ematch specific configuration data
349 * @data: ematch specific data
352 struct tcf_ematch_ops
* ops
;
354 unsigned int datalen
;
360 static inline int tcf_em_is_container(struct tcf_ematch
*em
)
365 static inline int tcf_em_is_simple(struct tcf_ematch
*em
)
367 return em
->flags
& TCF_EM_SIMPLE
;
370 static inline int tcf_em_is_inverted(struct tcf_ematch
*em
)
372 return em
->flags
& TCF_EM_INVERT
;
375 static inline int tcf_em_last_match(struct tcf_ematch
*em
)
377 return (em
->flags
& TCF_EM_REL_MASK
) == TCF_EM_REL_END
;
380 static inline int tcf_em_early_end(struct tcf_ematch
*em
, int result
)
382 if (tcf_em_last_match(em
))
385 if (result
== 0 && em
->flags
& TCF_EM_REL_AND
)
388 if (result
!= 0 && em
->flags
& TCF_EM_REL_OR
)
395 * struct tcf_ematch_tree - ematch tree handle
397 * @hdr: ematch tree header supplied by userspace
398 * @matches: array of ematches
400 struct tcf_ematch_tree
{
401 struct tcf_ematch_tree_hdr hdr
;
402 struct tcf_ematch
* matches
;
407 * struct tcf_ematch_ops - ematch module operations
409 * @kind: identifier (kind) of this ematch module
410 * @datalen: length of expected configuration data (optional)
411 * @change: called during validation (optional)
412 * @match: called during ematch tree evaluation, must return 1/0
413 * @destroy: called during destroyage (optional)
414 * @dump: called during dumping process (optional)
415 * @owner: owner, must be set to THIS_MODULE
416 * @link: link to previous/next ematch module (internal use)
418 struct tcf_ematch_ops
{
421 int (*change
)(struct net
*net
, void *,
422 int, struct tcf_ematch
*);
423 int (*match
)(struct sk_buff
*, struct tcf_ematch
*,
424 struct tcf_pkt_info
*);
425 void (*destroy
)(struct tcf_ematch
*);
426 int (*dump
)(struct sk_buff
*, struct tcf_ematch
*);
427 struct module
*owner
;
428 struct list_head link
;
431 int tcf_em_register(struct tcf_ematch_ops
*);
432 void tcf_em_unregister(struct tcf_ematch_ops
*);
433 int tcf_em_tree_validate(struct tcf_proto
*, struct nlattr
*,
434 struct tcf_ematch_tree
*);
435 void tcf_em_tree_destroy(struct tcf_ematch_tree
*);
436 int tcf_em_tree_dump(struct sk_buff
*, struct tcf_ematch_tree
*, int);
437 int __tcf_em_tree_match(struct sk_buff
*, struct tcf_ematch_tree
*,
438 struct tcf_pkt_info
*);
441 * tcf_em_tree_match - evaulate an ematch tree
443 * @skb: socket buffer of the packet in question
444 * @tree: ematch tree to be used for evaluation
445 * @info: packet information examined by classifier
447 * This function matches @skb against the ematch tree in @tree by going
448 * through all ematches respecting their logic relations returning
449 * as soon as the result is obvious.
451 * Returns 1 if the ematch tree as-one matches, no ematches are configured
452 * or ematch is not enabled in the kernel, otherwise 0 is returned.
454 static inline int tcf_em_tree_match(struct sk_buff
*skb
,
455 struct tcf_ematch_tree
*tree
,
456 struct tcf_pkt_info
*info
)
458 if (tree
->hdr
.nmatches
)
459 return __tcf_em_tree_match(skb
, tree
, info
);
464 #define MODULE_ALIAS_TCF_EMATCH(kind) MODULE_ALIAS("ematch-kind-" __stringify(kind))
466 #else /* CONFIG_NET_EMATCH */
468 struct tcf_ematch_tree
{
471 #define tcf_em_tree_validate(tp, tb, t) ((void)(t), 0)
472 #define tcf_em_tree_destroy(t) do { (void)(t); } while(0)
473 #define tcf_em_tree_dump(skb, t, tlv) (0)
474 #define tcf_em_tree_match(skb, t, info) ((void)(info), 1)
476 #endif /* CONFIG_NET_EMATCH */
478 static inline unsigned char * tcf_get_base_ptr(struct sk_buff
*skb
, int layer
)
482 return skb_mac_header(skb
);
483 case TCF_LAYER_NETWORK
:
484 return skb_network_header(skb
);
485 case TCF_LAYER_TRANSPORT
:
486 return skb_transport_header(skb
);
492 static inline int tcf_valid_offset(const struct sk_buff
*skb
,
493 const unsigned char *ptr
, const int len
)
495 return likely((ptr
+ len
) <= skb_tail_pointer(skb
) &&
497 (ptr
<= (ptr
+ len
)));
501 tcf_change_indev(struct net
*net
, struct nlattr
*indev_tlv
,
502 struct netlink_ext_ack
*extack
)
504 char indev
[IFNAMSIZ
];
505 struct net_device
*dev
;
507 if (nla_strlcpy(indev
, indev_tlv
, IFNAMSIZ
) >= IFNAMSIZ
) {
508 NL_SET_ERR_MSG_ATTR(extack
, indev_tlv
,
509 "Interface name too long");
512 dev
= __dev_get_by_name(net
, indev
);
514 NL_SET_ERR_MSG_ATTR(extack
, indev_tlv
,
515 "Network device not found");
522 tcf_match_indev(struct sk_buff
*skb
, int ifindex
)
528 return ifindex
== skb
->skb_iif
;
531 int tc_setup_flow_action(struct flow_action
*flow_action
,
532 const struct tcf_exts
*exts
);
533 void tc_cleanup_flow_action(struct flow_action
*flow_action
);
535 int tc_setup_cb_call(struct tcf_block
*block
, enum tc_setup_type type
,
536 void *type_data
, bool err_stop
, bool rtnl_held
);
537 int tc_setup_cb_add(struct tcf_block
*block
, struct tcf_proto
*tp
,
538 enum tc_setup_type type
, void *type_data
, bool err_stop
,
539 u32
*flags
, unsigned int *in_hw_count
, bool rtnl_held
);
540 int tc_setup_cb_replace(struct tcf_block
*block
, struct tcf_proto
*tp
,
541 enum tc_setup_type type
, void *type_data
, bool err_stop
,
542 u32
*old_flags
, unsigned int *old_in_hw_count
,
543 u32
*new_flags
, unsigned int *new_in_hw_count
,
545 int tc_setup_cb_destroy(struct tcf_block
*block
, struct tcf_proto
*tp
,
546 enum tc_setup_type type
, void *type_data
, bool err_stop
,
547 u32
*flags
, unsigned int *in_hw_count
, bool rtnl_held
);
548 int tc_setup_cb_reoffload(struct tcf_block
*block
, struct tcf_proto
*tp
,
549 bool add
, flow_setup_cb_t
*cb
,
550 enum tc_setup_type type
, void *type_data
,
551 void *cb_priv
, u32
*flags
, unsigned int *in_hw_count
);
552 unsigned int tcf_exts_num_actions(struct tcf_exts
*exts
);
554 struct tc_cls_u32_knode
{
555 struct tcf_exts
*exts
;
556 struct tcf_result
*res
;
557 struct tc_u32_sel
*sel
;
565 struct tc_cls_u32_hnode
{
568 unsigned int divisor
;
571 enum tc_clsu32_command
{
573 TC_CLSU32_REPLACE_KNODE
,
574 TC_CLSU32_DELETE_KNODE
,
576 TC_CLSU32_REPLACE_HNODE
,
577 TC_CLSU32_DELETE_HNODE
,
580 struct tc_cls_u32_offload
{
581 struct flow_cls_common_offload common
;
583 enum tc_clsu32_command command
;
585 struct tc_cls_u32_knode knode
;
586 struct tc_cls_u32_hnode hnode
;
590 static inline bool tc_can_offload(const struct net_device
*dev
)
592 return dev
->features
& NETIF_F_HW_TC
;
595 static inline bool tc_can_offload_extack(const struct net_device
*dev
,
596 struct netlink_ext_ack
*extack
)
598 bool can
= tc_can_offload(dev
);
601 NL_SET_ERR_MSG(extack
, "TC offload is disabled on net device");
607 tc_cls_can_offload_and_chain0(const struct net_device
*dev
,
608 struct flow_cls_common_offload
*common
)
610 if (!tc_can_offload_extack(dev
, common
->extack
))
612 if (common
->chain_index
) {
613 NL_SET_ERR_MSG(common
->extack
,
614 "Driver supports only offload of chain 0");
620 static inline bool tc_skip_hw(u32 flags
)
622 return (flags
& TCA_CLS_FLAGS_SKIP_HW
) ? true : false;
625 static inline bool tc_skip_sw(u32 flags
)
627 return (flags
& TCA_CLS_FLAGS_SKIP_SW
) ? true : false;
630 /* SKIP_HW and SKIP_SW are mutually exclusive flags. */
631 static inline bool tc_flags_valid(u32 flags
)
633 if (flags
& ~(TCA_CLS_FLAGS_SKIP_HW
| TCA_CLS_FLAGS_SKIP_SW
|
634 TCA_CLS_FLAGS_VERBOSE
))
637 flags
&= TCA_CLS_FLAGS_SKIP_HW
| TCA_CLS_FLAGS_SKIP_SW
;
638 if (!(flags
^ (TCA_CLS_FLAGS_SKIP_HW
| TCA_CLS_FLAGS_SKIP_SW
)))
644 static inline bool tc_in_hw(u32 flags
)
646 return (flags
& TCA_CLS_FLAGS_IN_HW
) ? true : false;
650 tc_cls_common_offload_init(struct flow_cls_common_offload
*cls_common
,
651 const struct tcf_proto
*tp
, u32 flags
,
652 struct netlink_ext_ack
*extack
)
654 cls_common
->chain_index
= tp
->chain
->index
;
655 cls_common
->protocol
= tp
->protocol
;
656 cls_common
->prio
= tp
->prio
>> 16;
657 if (tc_skip_sw(flags
) || flags
& TCA_CLS_FLAGS_VERBOSE
)
658 cls_common
->extack
= extack
;
661 enum tc_matchall_command
{
662 TC_CLSMATCHALL_REPLACE
,
663 TC_CLSMATCHALL_DESTROY
,
664 TC_CLSMATCHALL_STATS
,
667 struct tc_cls_matchall_offload
{
668 struct flow_cls_common_offload common
;
669 enum tc_matchall_command command
;
670 struct flow_rule
*rule
;
671 struct flow_stats stats
;
672 unsigned long cookie
;
675 enum tc_clsbpf_command
{
680 struct tc_cls_bpf_offload
{
681 struct flow_cls_common_offload common
;
682 enum tc_clsbpf_command command
;
683 struct tcf_exts
*exts
;
684 struct bpf_prog
*prog
;
685 struct bpf_prog
*oldprog
;
687 bool exts_integrated
;
690 struct tc_mqprio_qopt_offload
{
691 /* struct tc_mqprio_qopt must always be the first element */
692 struct tc_mqprio_qopt qopt
;
696 u64 min_rate
[TC_QOPT_MAX_QUEUE
];
697 u64 max_rate
[TC_QOPT_MAX_QUEUE
];
700 /* This structure holds cookie structure that is passed from user
701 * to the kernel for actions and classifiers
709 struct tc_qopt_offload_stats
{
710 struct gnet_stats_basic_packed
*bstats
;
711 struct gnet_stats_queue
*qstats
;
721 struct tc_mq_opt_offload_graft_params
{
726 struct tc_mq_qopt_offload
{
727 enum tc_mq_command command
;
730 struct tc_qopt_offload_stats stats
;
731 struct tc_mq_opt_offload_graft_params graft_params
;
735 enum tc_red_command
{
743 struct tc_red_qopt_offload_params
{
751 struct gnet_stats_queue
*qstats
;
754 struct tc_red_qopt_offload
{
755 enum tc_red_command command
;
759 struct tc_red_qopt_offload_params set
;
760 struct tc_qopt_offload_stats stats
;
761 struct red_stats
*xstats
;
766 enum tc_gred_command
{
772 struct tc_gred_vq_qopt_offload_params
{
781 /* Only need backlog, see struct tc_prio_qopt_offload_params */
785 struct tc_gred_qopt_offload_params
{
790 struct gnet_stats_queue
*qstats
;
791 struct tc_gred_vq_qopt_offload_params tab
[MAX_DPs
];
794 struct tc_gred_qopt_offload_stats
{
795 struct gnet_stats_basic_packed bstats
[MAX_DPs
];
796 struct gnet_stats_queue qstats
[MAX_DPs
];
797 struct red_stats
*xstats
[MAX_DPs
];
800 struct tc_gred_qopt_offload
{
801 enum tc_gred_command command
;
805 struct tc_gred_qopt_offload_params set
;
806 struct tc_gred_qopt_offload_stats stats
;
810 enum tc_prio_command
{
817 struct tc_prio_qopt_offload_params
{
819 u8 priomap
[TC_PRIO_MAX
+ 1];
820 /* At the point of un-offloading the Qdisc, the reported backlog and
821 * qlen need to be reduced by the portion that is in HW.
823 struct gnet_stats_queue
*qstats
;
826 struct tc_prio_qopt_offload_graft_params
{
831 struct tc_prio_qopt_offload
{
832 enum tc_prio_command command
;
836 struct tc_prio_qopt_offload_params replace_params
;
837 struct tc_qopt_offload_stats stats
;
838 struct tc_prio_qopt_offload_graft_params graft_params
;
842 enum tc_root_command
{
846 struct tc_root_qopt_offload
{
847 enum tc_root_command command
;
852 enum tc_ets_command
{
859 struct tc_ets_qopt_offload_replace_params
{
861 u8 priomap
[TC_PRIO_MAX
+ 1];
862 unsigned int quanta
[TCQ_ETS_MAX_BANDS
]; /* 0 for strict bands. */
863 unsigned int weights
[TCQ_ETS_MAX_BANDS
];
864 struct gnet_stats_queue
*qstats
;
867 struct tc_ets_qopt_offload_graft_params
{
872 struct tc_ets_qopt_offload
{
873 enum tc_ets_command command
;
877 struct tc_ets_qopt_offload_replace_params replace_params
;
878 struct tc_qopt_offload_stats stats
;
879 struct tc_ets_qopt_offload_graft_params graft_params
;
883 enum tc_tbf_command
{
889 struct tc_tbf_qopt_offload_replace_params
{
890 struct psched_ratecfg rate
;
892 struct gnet_stats_queue
*qstats
;
895 struct tc_tbf_qopt_offload
{
896 enum tc_tbf_command command
;
900 struct tc_tbf_qopt_offload_replace_params replace_params
;
901 struct tc_qopt_offload_stats stats
;
905 enum tc_fifo_command
{
911 struct tc_fifo_qopt_offload
{
912 enum tc_fifo_command command
;
916 struct tc_qopt_offload_stats stats
;