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
);
77 static inline bool tcf_block_shared(struct tcf_block
*block
)
82 static inline bool tcf_block_non_null_shared(struct tcf_block
*block
)
88 int tcf_block_get(struct tcf_block
**p_block
,
89 struct tcf_proto __rcu
**p_filter_chain
, struct Qdisc
*q
,
90 struct netlink_ext_ack
*extack
)
96 int tcf_block_get_ext(struct tcf_block
**p_block
, struct Qdisc
*q
,
97 struct tcf_block_ext_info
*ei
,
98 struct netlink_ext_ack
*extack
)
103 static inline void tcf_block_put(struct tcf_block
*block
)
108 void tcf_block_put_ext(struct tcf_block
*block
, struct Qdisc
*q
,
109 struct tcf_block_ext_info
*ei
)
113 static inline struct Qdisc
*tcf_block_q(struct tcf_block
*block
)
119 int tc_setup_cb_block_register(struct tcf_block
*block
, flow_setup_cb_t
*cb
,
126 void tc_setup_cb_block_unregister(struct tcf_block
*block
, flow_setup_cb_t
*cb
,
131 static inline int tcf_classify(struct sk_buff
*skb
, const struct tcf_proto
*tp
,
132 struct tcf_result
*res
, bool compat_mode
)
134 return TC_ACT_UNSPEC
;
138 static inline unsigned long
139 __cls_set_class(unsigned long *clp
, unsigned long cl
)
141 return xchg(clp
, cl
);
145 __tcf_bind_filter(struct Qdisc
*q
, struct tcf_result
*r
, unsigned long base
)
149 cl
= q
->ops
->cl_ops
->bind_tcf(q
, base
, r
->classid
);
150 cl
= __cls_set_class(&r
->class, cl
);
152 q
->ops
->cl_ops
->unbind_tcf(q
, cl
);
156 tcf_bind_filter(struct tcf_proto
*tp
, struct tcf_result
*r
, unsigned long base
)
158 struct Qdisc
*q
= tp
->chain
->block
->q
;
160 /* Check q as it is not set for shared blocks. In that case,
161 * setting class is not supported.
166 __tcf_bind_filter(q
, r
, base
);
171 __tcf_unbind_filter(struct Qdisc
*q
, struct tcf_result
*r
)
175 if ((cl
= __cls_set_class(&r
->class, 0)) != 0)
176 q
->ops
->cl_ops
->unbind_tcf(q
, cl
);
180 tcf_unbind_filter(struct tcf_proto
*tp
, struct tcf_result
*r
)
182 struct Qdisc
*q
= tp
->chain
->block
->q
;
186 __tcf_unbind_filter(q
, r
);
190 #ifdef CONFIG_NET_CLS_ACT
191 __u32 type
; /* for backward compat(TCA_OLD_COMPAT) */
193 struct tc_action
**actions
;
196 /* Map to export classifier specific extension TLV types to the
197 * generic extensions API. Unsupported extensions must be set to 0.
203 static inline int tcf_exts_init(struct tcf_exts
*exts
, struct net
*net
,
204 int action
, int police
)
206 #ifdef CONFIG_NET_CLS_ACT
208 exts
->nr_actions
= 0;
210 exts
->actions
= kcalloc(TCA_ACT_MAX_PRIO
, sizeof(struct tc_action
*),
215 exts
->action
= action
;
216 exts
->police
= police
;
220 /* Return false if the netns is being destroyed in cleanup_net(). Callers
221 * need to do cleanup synchronously in this case, otherwise may race with
222 * tc_action_net_exit(). Return true for other cases.
224 static inline bool tcf_exts_get_net(struct tcf_exts
*exts
)
226 #ifdef CONFIG_NET_CLS_ACT
227 exts
->net
= maybe_get_net(exts
->net
);
228 return exts
->net
!= NULL
;
234 static inline void tcf_exts_put_net(struct tcf_exts
*exts
)
236 #ifdef CONFIG_NET_CLS_ACT
242 #ifdef CONFIG_NET_CLS_ACT
243 #define tcf_exts_for_each_action(i, a, exts) \
244 for (i = 0; i < TCA_ACT_MAX_PRIO && ((a) = (exts)->actions[i]); i++)
246 #define tcf_exts_for_each_action(i, a, exts) \
247 for (; 0; (void)(i), (void)(a), (void)(exts))
251 tcf_exts_stats_update(const struct tcf_exts
*exts
,
252 u64 bytes
, u64 packets
, u64 lastuse
)
254 #ifdef CONFIG_NET_CLS_ACT
259 for (i
= 0; i
< exts
->nr_actions
; i
++) {
260 struct tc_action
*a
= exts
->actions
[i
];
262 tcf_action_stats_update(a
, bytes
, packets
, lastuse
, true);
270 * tcf_exts_has_actions - check if at least one action is present
271 * @exts: tc filter extensions handle
273 * Returns true if at least one action is present.
275 static inline bool tcf_exts_has_actions(struct tcf_exts
*exts
)
277 #ifdef CONFIG_NET_CLS_ACT
278 return exts
->nr_actions
;
285 * tcf_exts_exec - execute tc filter extensions
286 * @skb: socket buffer
287 * @exts: tc filter extensions handle
288 * @res: desired result
290 * Executes all configured extensions. Returns TC_ACT_OK on a normal execution,
291 * a negative number if the filter must be considered unmatched or
292 * a positive action code (TC_ACT_*) which must be returned to the
296 tcf_exts_exec(struct sk_buff
*skb
, struct tcf_exts
*exts
,
297 struct tcf_result
*res
)
299 #ifdef CONFIG_NET_CLS_ACT
300 return tcf_action_exec(skb
, exts
->actions
, exts
->nr_actions
, res
);
305 int tcf_exts_validate(struct net
*net
, struct tcf_proto
*tp
,
306 struct nlattr
**tb
, struct nlattr
*rate_tlv
,
307 struct tcf_exts
*exts
, bool ovr
, bool rtnl_held
,
308 struct netlink_ext_ack
*extack
);
309 void tcf_exts_destroy(struct tcf_exts
*exts
);
310 void tcf_exts_change(struct tcf_exts
*dst
, struct tcf_exts
*src
);
311 int tcf_exts_dump(struct sk_buff
*skb
, struct tcf_exts
*exts
);
312 int tcf_exts_dump_stats(struct sk_buff
*skb
, struct tcf_exts
*exts
);
315 * struct tcf_pkt_info - packet information
317 struct tcf_pkt_info
{
322 #ifdef CONFIG_NET_EMATCH
324 struct tcf_ematch_ops
;
327 * struct tcf_ematch - extended match (ematch)
329 * @matchid: identifier to allow userspace to reidentify a match
330 * @flags: flags specifying attributes and the relation to other matches
331 * @ops: the operations lookup table of the corresponding ematch module
332 * @datalen: length of the ematch specific configuration data
333 * @data: ematch specific data
336 struct tcf_ematch_ops
* ops
;
338 unsigned int datalen
;
344 static inline int tcf_em_is_container(struct tcf_ematch
*em
)
349 static inline int tcf_em_is_simple(struct tcf_ematch
*em
)
351 return em
->flags
& TCF_EM_SIMPLE
;
354 static inline int tcf_em_is_inverted(struct tcf_ematch
*em
)
356 return em
->flags
& TCF_EM_INVERT
;
359 static inline int tcf_em_last_match(struct tcf_ematch
*em
)
361 return (em
->flags
& TCF_EM_REL_MASK
) == TCF_EM_REL_END
;
364 static inline int tcf_em_early_end(struct tcf_ematch
*em
, int result
)
366 if (tcf_em_last_match(em
))
369 if (result
== 0 && em
->flags
& TCF_EM_REL_AND
)
372 if (result
!= 0 && em
->flags
& TCF_EM_REL_OR
)
379 * struct tcf_ematch_tree - ematch tree handle
381 * @hdr: ematch tree header supplied by userspace
382 * @matches: array of ematches
384 struct tcf_ematch_tree
{
385 struct tcf_ematch_tree_hdr hdr
;
386 struct tcf_ematch
* matches
;
391 * struct tcf_ematch_ops - ematch module operations
393 * @kind: identifier (kind) of this ematch module
394 * @datalen: length of expected configuration data (optional)
395 * @change: called during validation (optional)
396 * @match: called during ematch tree evaluation, must return 1/0
397 * @destroy: called during destroyage (optional)
398 * @dump: called during dumping process (optional)
399 * @owner: owner, must be set to THIS_MODULE
400 * @link: link to previous/next ematch module (internal use)
402 struct tcf_ematch_ops
{
405 int (*change
)(struct net
*net
, void *,
406 int, struct tcf_ematch
*);
407 int (*match
)(struct sk_buff
*, struct tcf_ematch
*,
408 struct tcf_pkt_info
*);
409 void (*destroy
)(struct tcf_ematch
*);
410 int (*dump
)(struct sk_buff
*, struct tcf_ematch
*);
411 struct module
*owner
;
412 struct list_head link
;
415 int tcf_em_register(struct tcf_ematch_ops
*);
416 void tcf_em_unregister(struct tcf_ematch_ops
*);
417 int tcf_em_tree_validate(struct tcf_proto
*, struct nlattr
*,
418 struct tcf_ematch_tree
*);
419 void tcf_em_tree_destroy(struct tcf_ematch_tree
*);
420 int tcf_em_tree_dump(struct sk_buff
*, struct tcf_ematch_tree
*, int);
421 int __tcf_em_tree_match(struct sk_buff
*, struct tcf_ematch_tree
*,
422 struct tcf_pkt_info
*);
425 * tcf_em_tree_match - evaulate an ematch tree
427 * @skb: socket buffer of the packet in question
428 * @tree: ematch tree to be used for evaluation
429 * @info: packet information examined by classifier
431 * This function matches @skb against the ematch tree in @tree by going
432 * through all ematches respecting their logic relations returning
433 * as soon as the result is obvious.
435 * Returns 1 if the ematch tree as-one matches, no ematches are configured
436 * or ematch is not enabled in the kernel, otherwise 0 is returned.
438 static inline int tcf_em_tree_match(struct sk_buff
*skb
,
439 struct tcf_ematch_tree
*tree
,
440 struct tcf_pkt_info
*info
)
442 if (tree
->hdr
.nmatches
)
443 return __tcf_em_tree_match(skb
, tree
, info
);
448 #define MODULE_ALIAS_TCF_EMATCH(kind) MODULE_ALIAS("ematch-kind-" __stringify(kind))
450 #else /* CONFIG_NET_EMATCH */
452 struct tcf_ematch_tree
{
455 #define tcf_em_tree_validate(tp, tb, t) ((void)(t), 0)
456 #define tcf_em_tree_destroy(t) do { (void)(t); } while(0)
457 #define tcf_em_tree_dump(skb, t, tlv) (0)
458 #define tcf_em_tree_match(skb, t, info) ((void)(info), 1)
460 #endif /* CONFIG_NET_EMATCH */
462 static inline unsigned char * tcf_get_base_ptr(struct sk_buff
*skb
, int layer
)
466 return skb_mac_header(skb
);
467 case TCF_LAYER_NETWORK
:
468 return skb_network_header(skb
);
469 case TCF_LAYER_TRANSPORT
:
470 return skb_transport_header(skb
);
476 static inline int tcf_valid_offset(const struct sk_buff
*skb
,
477 const unsigned char *ptr
, const int len
)
479 return likely((ptr
+ len
) <= skb_tail_pointer(skb
) &&
481 (ptr
<= (ptr
+ len
)));
485 tcf_change_indev(struct net
*net
, struct nlattr
*indev_tlv
,
486 struct netlink_ext_ack
*extack
)
488 char indev
[IFNAMSIZ
];
489 struct net_device
*dev
;
491 if (nla_strlcpy(indev
, indev_tlv
, IFNAMSIZ
) >= IFNAMSIZ
) {
492 NL_SET_ERR_MSG(extack
, "Interface name too long");
495 dev
= __dev_get_by_name(net
, indev
);
502 tcf_match_indev(struct sk_buff
*skb
, int ifindex
)
508 return ifindex
== skb
->skb_iif
;
511 int tc_setup_flow_action(struct flow_action
*flow_action
,
512 const struct tcf_exts
*exts
, bool rtnl_held
);
513 void tc_cleanup_flow_action(struct flow_action
*flow_action
);
515 int tc_setup_cb_call(struct tcf_block
*block
, enum tc_setup_type type
,
516 void *type_data
, bool err_stop
, bool rtnl_held
);
517 int tc_setup_cb_add(struct tcf_block
*block
, struct tcf_proto
*tp
,
518 enum tc_setup_type type
, void *type_data
, bool err_stop
,
519 u32
*flags
, unsigned int *in_hw_count
, bool rtnl_held
);
520 int tc_setup_cb_replace(struct tcf_block
*block
, struct tcf_proto
*tp
,
521 enum tc_setup_type type
, void *type_data
, bool err_stop
,
522 u32
*old_flags
, unsigned int *old_in_hw_count
,
523 u32
*new_flags
, unsigned int *new_in_hw_count
,
525 int tc_setup_cb_destroy(struct tcf_block
*block
, struct tcf_proto
*tp
,
526 enum tc_setup_type type
, void *type_data
, bool err_stop
,
527 u32
*flags
, unsigned int *in_hw_count
, bool rtnl_held
);
528 int tc_setup_cb_reoffload(struct tcf_block
*block
, struct tcf_proto
*tp
,
529 bool add
, flow_setup_cb_t
*cb
,
530 enum tc_setup_type type
, void *type_data
,
531 void *cb_priv
, u32
*flags
, unsigned int *in_hw_count
);
532 unsigned int tcf_exts_num_actions(struct tcf_exts
*exts
);
534 struct tc_cls_u32_knode
{
535 struct tcf_exts
*exts
;
536 struct tcf_result
*res
;
537 struct tc_u32_sel
*sel
;
545 struct tc_cls_u32_hnode
{
548 unsigned int divisor
;
551 enum tc_clsu32_command
{
553 TC_CLSU32_REPLACE_KNODE
,
554 TC_CLSU32_DELETE_KNODE
,
556 TC_CLSU32_REPLACE_HNODE
,
557 TC_CLSU32_DELETE_HNODE
,
560 struct tc_cls_u32_offload
{
561 struct flow_cls_common_offload common
;
563 enum tc_clsu32_command command
;
565 struct tc_cls_u32_knode knode
;
566 struct tc_cls_u32_hnode hnode
;
570 static inline bool tc_can_offload(const struct net_device
*dev
)
572 return dev
->features
& NETIF_F_HW_TC
;
575 static inline bool tc_can_offload_extack(const struct net_device
*dev
,
576 struct netlink_ext_ack
*extack
)
578 bool can
= tc_can_offload(dev
);
581 NL_SET_ERR_MSG(extack
, "TC offload is disabled on net device");
587 tc_cls_can_offload_and_chain0(const struct net_device
*dev
,
588 struct flow_cls_common_offload
*common
)
590 if (!tc_can_offload_extack(dev
, common
->extack
))
592 if (common
->chain_index
) {
593 NL_SET_ERR_MSG(common
->extack
,
594 "Driver supports only offload of chain 0");
600 static inline bool tc_skip_hw(u32 flags
)
602 return (flags
& TCA_CLS_FLAGS_SKIP_HW
) ? true : false;
605 static inline bool tc_skip_sw(u32 flags
)
607 return (flags
& TCA_CLS_FLAGS_SKIP_SW
) ? true : false;
610 /* SKIP_HW and SKIP_SW are mutually exclusive flags. */
611 static inline bool tc_flags_valid(u32 flags
)
613 if (flags
& ~(TCA_CLS_FLAGS_SKIP_HW
| TCA_CLS_FLAGS_SKIP_SW
|
614 TCA_CLS_FLAGS_VERBOSE
))
617 flags
&= TCA_CLS_FLAGS_SKIP_HW
| TCA_CLS_FLAGS_SKIP_SW
;
618 if (!(flags
^ (TCA_CLS_FLAGS_SKIP_HW
| TCA_CLS_FLAGS_SKIP_SW
)))
624 static inline bool tc_in_hw(u32 flags
)
626 return (flags
& TCA_CLS_FLAGS_IN_HW
) ? true : false;
630 tc_cls_common_offload_init(struct flow_cls_common_offload
*cls_common
,
631 const struct tcf_proto
*tp
, u32 flags
,
632 struct netlink_ext_ack
*extack
)
634 cls_common
->chain_index
= tp
->chain
->index
;
635 cls_common
->protocol
= tp
->protocol
;
636 cls_common
->prio
= tp
->prio
>> 16;
637 if (tc_skip_sw(flags
) || flags
& TCA_CLS_FLAGS_VERBOSE
)
638 cls_common
->extack
= extack
;
641 enum tc_matchall_command
{
642 TC_CLSMATCHALL_REPLACE
,
643 TC_CLSMATCHALL_DESTROY
,
644 TC_CLSMATCHALL_STATS
,
647 struct tc_cls_matchall_offload
{
648 struct flow_cls_common_offload common
;
649 enum tc_matchall_command command
;
650 struct flow_rule
*rule
;
651 struct flow_stats stats
;
652 unsigned long cookie
;
655 enum tc_clsbpf_command
{
660 struct tc_cls_bpf_offload
{
661 struct flow_cls_common_offload common
;
662 enum tc_clsbpf_command command
;
663 struct tcf_exts
*exts
;
664 struct bpf_prog
*prog
;
665 struct bpf_prog
*oldprog
;
667 bool exts_integrated
;
670 struct tc_mqprio_qopt_offload
{
671 /* struct tc_mqprio_qopt must always be the first element */
672 struct tc_mqprio_qopt qopt
;
676 u64 min_rate
[TC_QOPT_MAX_QUEUE
];
677 u64 max_rate
[TC_QOPT_MAX_QUEUE
];
680 /* This structure holds cookie structure that is passed from user
681 * to the kernel for actions and classifiers
689 struct tc_qopt_offload_stats
{
690 struct gnet_stats_basic_packed
*bstats
;
691 struct gnet_stats_queue
*qstats
;
701 struct tc_mq_opt_offload_graft_params
{
706 struct tc_mq_qopt_offload
{
707 enum tc_mq_command command
;
710 struct tc_qopt_offload_stats stats
;
711 struct tc_mq_opt_offload_graft_params graft_params
;
715 enum tc_red_command
{
723 struct tc_red_qopt_offload_params
{
730 struct gnet_stats_queue
*qstats
;
733 struct tc_red_qopt_offload
{
734 enum tc_red_command command
;
738 struct tc_red_qopt_offload_params set
;
739 struct tc_qopt_offload_stats stats
;
740 struct red_stats
*xstats
;
745 enum tc_gred_command
{
751 struct tc_gred_vq_qopt_offload_params
{
760 /* Only need backlog, see struct tc_prio_qopt_offload_params */
764 struct tc_gred_qopt_offload_params
{
769 struct gnet_stats_queue
*qstats
;
770 struct tc_gred_vq_qopt_offload_params tab
[MAX_DPs
];
773 struct tc_gred_qopt_offload_stats
{
774 struct gnet_stats_basic_packed bstats
[MAX_DPs
];
775 struct gnet_stats_queue qstats
[MAX_DPs
];
776 struct red_stats
*xstats
[MAX_DPs
];
779 struct tc_gred_qopt_offload
{
780 enum tc_gred_command command
;
784 struct tc_gred_qopt_offload_params set
;
785 struct tc_gred_qopt_offload_stats stats
;
789 enum tc_prio_command
{
796 struct tc_prio_qopt_offload_params
{
798 u8 priomap
[TC_PRIO_MAX
+ 1];
799 /* At the point of un-offloading the Qdisc, the reported backlog and
800 * qlen need to be reduced by the portion that is in HW.
802 struct gnet_stats_queue
*qstats
;
805 struct tc_prio_qopt_offload_graft_params
{
810 struct tc_prio_qopt_offload
{
811 enum tc_prio_command command
;
815 struct tc_prio_qopt_offload_params replace_params
;
816 struct tc_qopt_offload_stats stats
;
817 struct tc_prio_qopt_offload_graft_params graft_params
;
821 enum tc_root_command
{
825 struct tc_root_qopt_offload
{
826 enum tc_root_command command
;
831 enum tc_ets_command
{
838 struct tc_ets_qopt_offload_replace_params
{
840 u8 priomap
[TC_PRIO_MAX
+ 1];
841 unsigned int quanta
[TCQ_ETS_MAX_BANDS
]; /* 0 for strict bands. */
842 unsigned int weights
[TCQ_ETS_MAX_BANDS
];
843 struct gnet_stats_queue
*qstats
;
846 struct tc_ets_qopt_offload_graft_params
{
851 struct tc_ets_qopt_offload
{
852 enum tc_ets_command command
;
856 struct tc_ets_qopt_offload_replace_params replace_params
;
857 struct tc_qopt_offload_stats stats
;
858 struct tc_ets_qopt_offload_graft_params graft_params
;
862 enum tc_tbf_command
{
868 struct tc_tbf_qopt_offload_replace_params
{
869 struct psched_ratecfg rate
;
871 struct gnet_stats_queue
*qstats
;
874 struct tc_tbf_qopt_offload
{
875 enum tc_tbf_command command
;
879 struct tc_tbf_qopt_offload_replace_params replace_params
;
880 struct tc_qopt_offload_stats stats
;