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>
10 /* Basic packet classifier frontend definitions. */
16 int (*fn)(struct tcf_proto *, void *node, struct tcf_walker *);
19 int register_tcf_proto_ops(struct tcf_proto_ops *ops);
20 int unregister_tcf_proto_ops(struct tcf_proto_ops *ops);
22 enum tcf_block_binder_type {
23 TCF_BLOCK_BINDER_TYPE_UNSPEC,
24 TCF_BLOCK_BINDER_TYPE_CLSACT_INGRESS,
25 TCF_BLOCK_BINDER_TYPE_CLSACT_EGRESS,
28 struct tcf_block_ext_info {
29 enum tcf_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(struct tcf_block *block, u32 chain_index,
41 void tcf_chain_put(struct tcf_chain *chain);
42 void tcf_block_netif_keep_dst(struct tcf_block *block);
43 int tcf_block_get(struct tcf_block **p_block,
44 struct tcf_proto __rcu **p_filter_chain, struct Qdisc *q,
45 struct netlink_ext_ack *extack);
46 int tcf_block_get_ext(struct tcf_block **p_block, struct Qdisc *q,
47 struct tcf_block_ext_info *ei,
48 struct netlink_ext_ack *extack);
49 void tcf_block_put(struct tcf_block *block);
50 void tcf_block_put_ext(struct tcf_block *block, struct Qdisc *q,
51 struct tcf_block_ext_info *ei);
53 static inline bool tcf_block_shared(struct tcf_block *block)
58 static inline struct Qdisc *tcf_block_q(struct tcf_block *block)
60 WARN_ON(tcf_block_shared(block));
64 static inline struct net_device *tcf_block_dev(struct tcf_block *block)
66 return tcf_block_q(block)->dev_queue->dev;
69 void *tcf_block_cb_priv(struct tcf_block_cb *block_cb);
70 struct tcf_block_cb *tcf_block_cb_lookup(struct tcf_block *block,
71 tc_setup_cb_t *cb, void *cb_ident);
72 void tcf_block_cb_incref(struct tcf_block_cb *block_cb);
73 unsigned int tcf_block_cb_decref(struct tcf_block_cb *block_cb);
74 struct tcf_block_cb *__tcf_block_cb_register(struct tcf_block *block,
75 tc_setup_cb_t *cb, void *cb_ident,
77 int tcf_block_cb_register(struct tcf_block *block,
78 tc_setup_cb_t *cb, void *cb_ident,
80 void __tcf_block_cb_unregister(struct tcf_block_cb *block_cb);
81 void tcf_block_cb_unregister(struct tcf_block *block,
82 tc_setup_cb_t *cb, void *cb_ident);
84 int tcf_classify(struct sk_buff *skb, const struct tcf_proto *tp,
85 struct tcf_result *res, bool compat_mode);
89 int tcf_block_get(struct tcf_block **p_block,
90 struct tcf_proto __rcu **p_filter_chain, struct Qdisc *q,
91 struct netlink_ext_ack *extack)
97 int tcf_block_get_ext(struct tcf_block **p_block, struct Qdisc *q,
98 struct tcf_block_ext_info *ei,
99 struct netlink_ext_ack *extack)
104 static inline void tcf_block_put(struct tcf_block *block)
109 void tcf_block_put_ext(struct tcf_block *block, struct Qdisc *q,
110 struct tcf_block_ext_info *ei)
114 static inline bool tcf_block_shared(struct tcf_block *block)
119 static inline struct Qdisc *tcf_block_q(struct tcf_block *block)
124 static inline struct net_device *tcf_block_dev(struct tcf_block *block)
130 int tc_setup_cb_block_register(struct tcf_block *block, tc_setup_cb_t *cb,
137 void tc_setup_cb_block_unregister(struct tcf_block *block, tc_setup_cb_t *cb,
143 void *tcf_block_cb_priv(struct tcf_block_cb *block_cb)
149 struct tcf_block_cb *tcf_block_cb_lookup(struct tcf_block *block,
150 tc_setup_cb_t *cb, void *cb_ident)
156 void tcf_block_cb_incref(struct tcf_block_cb *block_cb)
161 unsigned int tcf_block_cb_decref(struct tcf_block_cb *block_cb)
167 struct tcf_block_cb *__tcf_block_cb_register(struct tcf_block *block,
168 tc_setup_cb_t *cb, void *cb_ident,
175 int tcf_block_cb_register(struct tcf_block *block,
176 tc_setup_cb_t *cb, void *cb_ident,
183 void __tcf_block_cb_unregister(struct tcf_block_cb *block_cb)
188 void tcf_block_cb_unregister(struct tcf_block *block,
189 tc_setup_cb_t *cb, void *cb_ident)
193 static inline int tcf_classify(struct sk_buff *skb, const struct tcf_proto *tp,
194 struct tcf_result *res, bool compat_mode)
196 return TC_ACT_UNSPEC;
200 static inline unsigned long
201 __cls_set_class(unsigned long *clp, unsigned long cl)
203 return xchg(clp, cl);
206 static inline unsigned long
207 cls_set_class(struct Qdisc *q, unsigned long *clp, unsigned long cl)
209 unsigned long old_cl;
212 old_cl = __cls_set_class(clp, cl);
218 tcf_bind_filter(struct tcf_proto *tp, struct tcf_result *r, unsigned long base)
220 struct Qdisc *q = tp->chain->block->q;
223 /* Check q as it is not set for shared blocks. In that case,
224 * setting class is not supported.
228 cl = q->ops->cl_ops->bind_tcf(q, base, r->classid);
229 cl = cls_set_class(q, &r->class, cl);
231 q->ops->cl_ops->unbind_tcf(q, cl);
235 tcf_unbind_filter(struct tcf_proto *tp, struct tcf_result *r)
237 struct Qdisc *q = tp->chain->block->q;
242 if ((cl = __cls_set_class(&r->class, 0)) != 0)
243 q->ops->cl_ops->unbind_tcf(q, cl);
247 #ifdef CONFIG_NET_CLS_ACT
248 __u32 type; /* for backward compat(TCA_OLD_COMPAT) */
250 struct tc_action **actions;
253 /* Map to export classifier specific extension TLV types to the
254 * generic extensions API. Unsupported extensions must be set to 0.
260 static inline int tcf_exts_init(struct tcf_exts *exts, int action, int police)
262 #ifdef CONFIG_NET_CLS_ACT
264 exts->nr_actions = 0;
266 exts->actions = kcalloc(TCA_ACT_MAX_PRIO, sizeof(struct tc_action *),
271 exts->action = action;
272 exts->police = police;
276 /* Return false if the netns is being destroyed in cleanup_net(). Callers
277 * need to do cleanup synchronously in this case, otherwise may race with
278 * tc_action_net_exit(). Return true for other cases.
280 static inline bool tcf_exts_get_net(struct tcf_exts *exts)
282 #ifdef CONFIG_NET_CLS_ACT
283 exts->net = maybe_get_net(exts->net);
284 return exts->net != NULL;
290 static inline void tcf_exts_put_net(struct tcf_exts *exts)
292 #ifdef CONFIG_NET_CLS_ACT
298 static inline void tcf_exts_to_list(const struct tcf_exts *exts,
299 struct list_head *actions)
301 #ifdef CONFIG_NET_CLS_ACT
304 for (i = 0; i < exts->nr_actions; i++) {
305 struct tc_action *a = exts->actions[i];
307 list_add_tail(&a->list, actions);
313 tcf_exts_stats_update(const struct tcf_exts *exts,
314 u64 bytes, u64 packets, u64 lastuse)
316 #ifdef CONFIG_NET_CLS_ACT
321 for (i = 0; i < exts->nr_actions; i++) {
322 struct tc_action *a = exts->actions[i];
324 tcf_action_stats_update(a, bytes, packets, lastuse);
332 * tcf_exts_has_actions - check if at least one action is present
333 * @exts: tc filter extensions handle
335 * Returns true if at least one action is present.
337 static inline bool tcf_exts_has_actions(struct tcf_exts *exts)
339 #ifdef CONFIG_NET_CLS_ACT
340 return exts->nr_actions;
347 * tcf_exts_has_one_action - check if exactly one action is present
348 * @exts: tc filter extensions handle
350 * Returns true if exactly one action is present.
352 static inline bool tcf_exts_has_one_action(struct tcf_exts *exts)
354 #ifdef CONFIG_NET_CLS_ACT
355 return exts->nr_actions == 1;
362 * tcf_exts_exec - execute tc filter extensions
363 * @skb: socket buffer
364 * @exts: tc filter extensions handle
365 * @res: desired result
367 * Executes all configured extensions. Returns TC_ACT_OK on a normal execution,
368 * a negative number if the filter must be considered unmatched or
369 * a positive action code (TC_ACT_*) which must be returned to the
373 tcf_exts_exec(struct sk_buff *skb, struct tcf_exts *exts,
374 struct tcf_result *res)
376 #ifdef CONFIG_NET_CLS_ACT
377 return tcf_action_exec(skb, exts->actions, exts->nr_actions, res);
382 int tcf_exts_validate(struct net *net, struct tcf_proto *tp,
383 struct nlattr **tb, struct nlattr *rate_tlv,
384 struct tcf_exts *exts, bool ovr,
385 struct netlink_ext_ack *extack);
386 void tcf_exts_destroy(struct tcf_exts *exts);
387 void tcf_exts_change(struct tcf_exts *dst, struct tcf_exts *src);
388 int tcf_exts_dump(struct sk_buff *skb, struct tcf_exts *exts);
389 int tcf_exts_dump_stats(struct sk_buff *skb, struct tcf_exts *exts);
392 * struct tcf_pkt_info - packet information
394 struct tcf_pkt_info {
399 #ifdef CONFIG_NET_EMATCH
401 struct tcf_ematch_ops;
404 * struct tcf_ematch - extended match (ematch)
406 * @matchid: identifier to allow userspace to reidentify a match
407 * @flags: flags specifying attributes and the relation to other matches
408 * @ops: the operations lookup table of the corresponding ematch module
409 * @datalen: length of the ematch specific configuration data
410 * @data: ematch specific data
413 struct tcf_ematch_ops * ops;
415 unsigned int datalen;
421 static inline int tcf_em_is_container(struct tcf_ematch *em)
426 static inline int tcf_em_is_simple(struct tcf_ematch *em)
428 return em->flags & TCF_EM_SIMPLE;
431 static inline int tcf_em_is_inverted(struct tcf_ematch *em)
433 return em->flags & TCF_EM_INVERT;
436 static inline int tcf_em_last_match(struct tcf_ematch *em)
438 return (em->flags & TCF_EM_REL_MASK) == TCF_EM_REL_END;
441 static inline int tcf_em_early_end(struct tcf_ematch *em, int result)
443 if (tcf_em_last_match(em))
446 if (result == 0 && em->flags & TCF_EM_REL_AND)
449 if (result != 0 && em->flags & TCF_EM_REL_OR)
456 * struct tcf_ematch_tree - ematch tree handle
458 * @hdr: ematch tree header supplied by userspace
459 * @matches: array of ematches
461 struct tcf_ematch_tree {
462 struct tcf_ematch_tree_hdr hdr;
463 struct tcf_ematch * matches;
468 * struct tcf_ematch_ops - ematch module operations
470 * @kind: identifier (kind) of this ematch module
471 * @datalen: length of expected configuration data (optional)
472 * @change: called during validation (optional)
473 * @match: called during ematch tree evaluation, must return 1/0
474 * @destroy: called during destroyage (optional)
475 * @dump: called during dumping process (optional)
476 * @owner: owner, must be set to THIS_MODULE
477 * @link: link to previous/next ematch module (internal use)
479 struct tcf_ematch_ops {
482 int (*change)(struct net *net, void *,
483 int, struct tcf_ematch *);
484 int (*match)(struct sk_buff *, struct tcf_ematch *,
485 struct tcf_pkt_info *);
486 void (*destroy)(struct tcf_ematch *);
487 int (*dump)(struct sk_buff *, struct tcf_ematch *);
488 struct module *owner;
489 struct list_head link;
492 int tcf_em_register(struct tcf_ematch_ops *);
493 void tcf_em_unregister(struct tcf_ematch_ops *);
494 int tcf_em_tree_validate(struct tcf_proto *, struct nlattr *,
495 struct tcf_ematch_tree *);
496 void tcf_em_tree_destroy(struct tcf_ematch_tree *);
497 int tcf_em_tree_dump(struct sk_buff *, struct tcf_ematch_tree *, int);
498 int __tcf_em_tree_match(struct sk_buff *, struct tcf_ematch_tree *,
499 struct tcf_pkt_info *);
502 * tcf_em_tree_match - evaulate an ematch tree
504 * @skb: socket buffer of the packet in question
505 * @tree: ematch tree to be used for evaluation
506 * @info: packet information examined by classifier
508 * This function matches @skb against the ematch tree in @tree by going
509 * through all ematches respecting their logic relations returning
510 * as soon as the result is obvious.
512 * Returns 1 if the ematch tree as-one matches, no ematches are configured
513 * or ematch is not enabled in the kernel, otherwise 0 is returned.
515 static inline int tcf_em_tree_match(struct sk_buff *skb,
516 struct tcf_ematch_tree *tree,
517 struct tcf_pkt_info *info)
519 if (tree->hdr.nmatches)
520 return __tcf_em_tree_match(skb, tree, info);
525 #define MODULE_ALIAS_TCF_EMATCH(kind) MODULE_ALIAS("ematch-kind-" __stringify(kind))
527 #else /* CONFIG_NET_EMATCH */
529 struct tcf_ematch_tree {
532 #define tcf_em_tree_validate(tp, tb, t) ((void)(t), 0)
533 #define tcf_em_tree_destroy(t) do { (void)(t); } while(0)
534 #define tcf_em_tree_dump(skb, t, tlv) (0)
535 #define tcf_em_tree_match(skb, t, info) ((void)(info), 1)
537 #endif /* CONFIG_NET_EMATCH */
539 static inline unsigned char * tcf_get_base_ptr(struct sk_buff *skb, int layer)
543 return skb_mac_header(skb);
544 case TCF_LAYER_NETWORK:
545 return skb_network_header(skb);
546 case TCF_LAYER_TRANSPORT:
547 return skb_transport_header(skb);
553 static inline int tcf_valid_offset(const struct sk_buff *skb,
554 const unsigned char *ptr, const int len)
556 return likely((ptr + len) <= skb_tail_pointer(skb) &&
558 (ptr <= (ptr + len)));
561 #ifdef CONFIG_NET_CLS_IND
562 #include <net/net_namespace.h>
565 tcf_change_indev(struct net *net, struct nlattr *indev_tlv,
566 struct netlink_ext_ack *extack)
568 char indev[IFNAMSIZ];
569 struct net_device *dev;
571 if (nla_strlcpy(indev, indev_tlv, IFNAMSIZ) >= IFNAMSIZ) {
572 NL_SET_ERR_MSG(extack, "Interface name too long");
575 dev = __dev_get_by_name(net, indev);
582 tcf_match_indev(struct sk_buff *skb, int ifindex)
588 return ifindex == skb->skb_iif;
590 #endif /* CONFIG_NET_CLS_IND */
592 int tc_setup_cb_call(struct tcf_block *block, struct tcf_exts *exts,
593 enum tc_setup_type type, void *type_data, bool err_stop);
595 enum tc_block_command {
600 struct tc_block_offload {
601 enum tc_block_command command;
602 enum tcf_block_binder_type binder_type;
603 struct tcf_block *block;
606 struct tc_cls_common_offload {
610 struct netlink_ext_ack *extack;
613 struct tc_cls_u32_knode {
614 struct tcf_exts *exts;
615 struct tc_u32_sel *sel;
623 struct tc_cls_u32_hnode {
626 unsigned int divisor;
629 enum tc_clsu32_command {
631 TC_CLSU32_REPLACE_KNODE,
632 TC_CLSU32_DELETE_KNODE,
634 TC_CLSU32_REPLACE_HNODE,
635 TC_CLSU32_DELETE_HNODE,
638 struct tc_cls_u32_offload {
639 struct tc_cls_common_offload common;
641 enum tc_clsu32_command command;
643 struct tc_cls_u32_knode knode;
644 struct tc_cls_u32_hnode hnode;
648 static inline bool tc_can_offload(const struct net_device *dev)
650 return dev->features & NETIF_F_HW_TC;
653 static inline bool tc_can_offload_extack(const struct net_device *dev,
654 struct netlink_ext_ack *extack)
656 bool can = tc_can_offload(dev);
659 NL_SET_ERR_MSG(extack, "TC offload is disabled on net device");
665 tc_cls_can_offload_and_chain0(const struct net_device *dev,
666 struct tc_cls_common_offload *common)
668 if (!tc_can_offload_extack(dev, common->extack))
670 if (common->chain_index) {
671 NL_SET_ERR_MSG(common->extack,
672 "Driver supports only offload of chain 0");
678 static inline bool tc_skip_hw(u32 flags)
680 return (flags & TCA_CLS_FLAGS_SKIP_HW) ? true : false;
683 static inline bool tc_skip_sw(u32 flags)
685 return (flags & TCA_CLS_FLAGS_SKIP_SW) ? true : false;
688 /* SKIP_HW and SKIP_SW are mutually exclusive flags. */
689 static inline bool tc_flags_valid(u32 flags)
691 if (flags & ~(TCA_CLS_FLAGS_SKIP_HW | TCA_CLS_FLAGS_SKIP_SW |
692 TCA_CLS_FLAGS_VERBOSE))
695 flags &= TCA_CLS_FLAGS_SKIP_HW | TCA_CLS_FLAGS_SKIP_SW;
696 if (!(flags ^ (TCA_CLS_FLAGS_SKIP_HW | TCA_CLS_FLAGS_SKIP_SW)))
702 static inline bool tc_in_hw(u32 flags)
704 return (flags & TCA_CLS_FLAGS_IN_HW) ? true : false;
708 tc_cls_common_offload_init(struct tc_cls_common_offload *cls_common,
709 const struct tcf_proto *tp, u32 flags,
710 struct netlink_ext_ack *extack)
712 cls_common->chain_index = tp->chain->index;
713 cls_common->protocol = tp->protocol;
714 cls_common->prio = tp->prio;
715 if (tc_skip_sw(flags) || flags & TCA_CLS_FLAGS_VERBOSE)
716 cls_common->extack = extack;
720 TC_CLSFLOWER_REPLACE,
721 TC_CLSFLOWER_DESTROY,
725 struct tc_cls_flower_offload {
726 struct tc_cls_common_offload common;
727 enum tc_fl_command command;
728 unsigned long cookie;
729 struct flow_dissector *dissector;
730 struct fl_flow_key *mask;
731 struct fl_flow_key *key;
732 struct tcf_exts *exts;
736 enum tc_matchall_command {
737 TC_CLSMATCHALL_REPLACE,
738 TC_CLSMATCHALL_DESTROY,
741 struct tc_cls_matchall_offload {
742 struct tc_cls_common_offload common;
743 enum tc_matchall_command command;
744 struct tcf_exts *exts;
745 unsigned long cookie;
748 enum tc_clsbpf_command {
753 struct tc_cls_bpf_offload {
754 struct tc_cls_common_offload common;
755 enum tc_clsbpf_command command;
756 struct tcf_exts *exts;
757 struct bpf_prog *prog;
758 struct bpf_prog *oldprog;
760 bool exts_integrated;
763 struct tc_mqprio_qopt_offload {
764 /* struct tc_mqprio_qopt must always be the first element */
765 struct tc_mqprio_qopt qopt;
769 u64 min_rate[TC_QOPT_MAX_QUEUE];
770 u64 max_rate[TC_QOPT_MAX_QUEUE];
773 /* This structure holds cookie structure that is passed from user
774 * to the kernel for actions and classifiers
781 struct tc_qopt_offload_stats {
782 struct gnet_stats_basic_packed *bstats;
783 struct gnet_stats_queue *qstats;
792 struct tc_mq_qopt_offload {
793 enum tc_mq_command command;
795 struct tc_qopt_offload_stats stats;
798 enum tc_red_command {
805 struct tc_red_qopt_offload_params {
810 struct gnet_stats_queue *qstats;
813 struct tc_red_qopt_offload {
814 enum tc_red_command command;
818 struct tc_red_qopt_offload_params set;
819 struct tc_qopt_offload_stats stats;
820 struct red_stats *xstats;
824 enum tc_prio_command {
831 struct tc_prio_qopt_offload_params {
833 u8 priomap[TC_PRIO_MAX + 1];
834 /* In case that a prio qdisc is offloaded and now is changed to a
835 * non-offloadedable config, it needs to update the backlog & qlen
836 * values to negate the HW backlog & qlen values (and only them).
838 struct gnet_stats_queue *qstats;
841 struct tc_prio_qopt_offload_graft_params {
846 struct tc_prio_qopt_offload {
847 enum tc_prio_command command;
851 struct tc_prio_qopt_offload_params replace_params;
852 struct tc_qopt_offload_stats stats;
853 struct tc_prio_qopt_offload_graft_params graft_params;