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/flow_offload.h>
11 /* TC action not accessible from user space */
12 #define TC_ACT_REINSERT (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 enum tcf_block_binder_type {
29 TCF_BLOCK_BINDER_TYPE_UNSPEC,
30 TCF_BLOCK_BINDER_TYPE_CLSACT_INGRESS,
31 TCF_BLOCK_BINDER_TYPE_CLSACT_EGRESS,
34 struct tcf_block_ext_info {
35 enum tcf_block_binder_type binder_type;
36 tcf_chain_head_change_t *chain_head_change;
37 void *chain_head_change_priv;
42 bool tcf_queue_work(struct rcu_work *rwork, work_func_t func);
45 struct tcf_chain *tcf_chain_get_by_act(struct tcf_block *block,
47 void tcf_chain_put_by_act(struct tcf_chain *chain);
48 struct tcf_chain *tcf_get_next_chain(struct tcf_block *block,
49 struct tcf_chain *chain);
50 struct tcf_proto *tcf_get_next_proto(struct tcf_chain *chain,
51 struct tcf_proto *tp, bool rtnl_held);
52 void tcf_block_netif_keep_dst(struct tcf_block *block);
53 int tcf_block_get(struct tcf_block **p_block,
54 struct tcf_proto __rcu **p_filter_chain, struct Qdisc *q,
55 struct netlink_ext_ack *extack);
56 int tcf_block_get_ext(struct tcf_block **p_block, struct Qdisc *q,
57 struct tcf_block_ext_info *ei,
58 struct netlink_ext_ack *extack);
59 void tcf_block_put(struct tcf_block *block);
60 void tcf_block_put_ext(struct tcf_block *block, struct Qdisc *q,
61 struct tcf_block_ext_info *ei);
63 static inline bool tcf_block_shared(struct tcf_block *block)
68 static inline struct Qdisc *tcf_block_q(struct tcf_block *block)
70 WARN_ON(tcf_block_shared(block));
74 void *tcf_block_cb_priv(struct tcf_block_cb *block_cb);
75 struct tcf_block_cb *tcf_block_cb_lookup(struct tcf_block *block,
76 tc_setup_cb_t *cb, void *cb_ident);
77 void tcf_block_cb_incref(struct tcf_block_cb *block_cb);
78 unsigned int tcf_block_cb_decref(struct tcf_block_cb *block_cb);
79 struct tcf_block_cb *__tcf_block_cb_register(struct tcf_block *block,
80 tc_setup_cb_t *cb, void *cb_ident,
82 struct netlink_ext_ack *extack);
83 int tcf_block_cb_register(struct tcf_block *block,
84 tc_setup_cb_t *cb, void *cb_ident,
85 void *cb_priv, struct netlink_ext_ack *extack);
86 void __tcf_block_cb_unregister(struct tcf_block *block,
87 struct tcf_block_cb *block_cb);
88 void tcf_block_cb_unregister(struct tcf_block *block,
89 tc_setup_cb_t *cb, void *cb_ident);
90 int __tc_indr_block_cb_register(struct net_device *dev, void *cb_priv,
91 tc_indr_block_bind_cb_t *cb, void *cb_ident);
92 int tc_indr_block_cb_register(struct net_device *dev, void *cb_priv,
93 tc_indr_block_bind_cb_t *cb, void *cb_ident);
94 void __tc_indr_block_cb_unregister(struct net_device *dev,
95 tc_indr_block_bind_cb_t *cb, void *cb_ident);
96 void tc_indr_block_cb_unregister(struct net_device *dev,
97 tc_indr_block_bind_cb_t *cb, void *cb_ident);
99 int tcf_classify(struct sk_buff *skb, const struct tcf_proto *tp,
100 struct tcf_result *res, bool compat_mode);
103 static inline bool tcf_block_shared(struct tcf_block *block)
109 int tcf_block_get(struct tcf_block **p_block,
110 struct tcf_proto __rcu **p_filter_chain, struct Qdisc *q,
111 struct netlink_ext_ack *extack)
117 int tcf_block_get_ext(struct tcf_block **p_block, struct Qdisc *q,
118 struct tcf_block_ext_info *ei,
119 struct netlink_ext_ack *extack)
124 static inline void tcf_block_put(struct tcf_block *block)
129 void tcf_block_put_ext(struct tcf_block *block, struct Qdisc *q,
130 struct tcf_block_ext_info *ei)
134 static inline struct Qdisc *tcf_block_q(struct tcf_block *block)
140 int tc_setup_cb_block_register(struct tcf_block *block, tc_setup_cb_t *cb,
147 void tc_setup_cb_block_unregister(struct tcf_block *block, tc_setup_cb_t *cb,
153 void *tcf_block_cb_priv(struct tcf_block_cb *block_cb)
159 struct tcf_block_cb *tcf_block_cb_lookup(struct tcf_block *block,
160 tc_setup_cb_t *cb, void *cb_ident)
166 void tcf_block_cb_incref(struct tcf_block_cb *block_cb)
171 unsigned int tcf_block_cb_decref(struct tcf_block_cb *block_cb)
177 struct tcf_block_cb *__tcf_block_cb_register(struct tcf_block *block,
178 tc_setup_cb_t *cb, void *cb_ident,
180 struct netlink_ext_ack *extack)
186 int tcf_block_cb_register(struct tcf_block *block,
187 tc_setup_cb_t *cb, void *cb_ident,
188 void *cb_priv, struct netlink_ext_ack *extack)
194 void __tcf_block_cb_unregister(struct tcf_block *block,
195 struct tcf_block_cb *block_cb)
200 void tcf_block_cb_unregister(struct tcf_block *block,
201 tc_setup_cb_t *cb, void *cb_ident)
206 int __tc_indr_block_cb_register(struct net_device *dev, void *cb_priv,
207 tc_indr_block_bind_cb_t *cb, void *cb_ident)
213 int tc_indr_block_cb_register(struct net_device *dev, void *cb_priv,
214 tc_indr_block_bind_cb_t *cb, void *cb_ident)
220 void __tc_indr_block_cb_unregister(struct net_device *dev,
221 tc_indr_block_bind_cb_t *cb, void *cb_ident)
226 void tc_indr_block_cb_unregister(struct net_device *dev,
227 tc_indr_block_bind_cb_t *cb, void *cb_ident)
231 static inline int tcf_classify(struct sk_buff *skb, const struct tcf_proto *tp,
232 struct tcf_result *res, bool compat_mode)
234 return TC_ACT_UNSPEC;
238 static inline unsigned long
239 __cls_set_class(unsigned long *clp, unsigned long cl)
241 return xchg(clp, cl);
244 static inline unsigned long
245 cls_set_class(struct Qdisc *q, unsigned long *clp, unsigned long cl)
247 unsigned long old_cl;
250 old_cl = __cls_set_class(clp, cl);
256 tcf_bind_filter(struct tcf_proto *tp, struct tcf_result *r, unsigned long base)
258 struct Qdisc *q = tp->chain->block->q;
261 /* Check q as it is not set for shared blocks. In that case,
262 * setting class is not supported.
266 cl = q->ops->cl_ops->bind_tcf(q, base, r->classid);
267 cl = cls_set_class(q, &r->class, cl);
269 q->ops->cl_ops->unbind_tcf(q, cl);
273 tcf_unbind_filter(struct tcf_proto *tp, struct tcf_result *r)
275 struct Qdisc *q = tp->chain->block->q;
280 if ((cl = __cls_set_class(&r->class, 0)) != 0)
281 q->ops->cl_ops->unbind_tcf(q, cl);
285 #ifdef CONFIG_NET_CLS_ACT
286 __u32 type; /* for backward compat(TCA_OLD_COMPAT) */
288 struct tc_action **actions;
291 /* Map to export classifier specific extension TLV types to the
292 * generic extensions API. Unsupported extensions must be set to 0.
298 static inline int tcf_exts_init(struct tcf_exts *exts, struct net *net,
299 int action, int police)
301 #ifdef CONFIG_NET_CLS_ACT
303 exts->nr_actions = 0;
305 exts->actions = kcalloc(TCA_ACT_MAX_PRIO, sizeof(struct tc_action *),
310 exts->action = action;
311 exts->police = police;
315 /* Return false if the netns is being destroyed in cleanup_net(). Callers
316 * need to do cleanup synchronously in this case, otherwise may race with
317 * tc_action_net_exit(). Return true for other cases.
319 static inline bool tcf_exts_get_net(struct tcf_exts *exts)
321 #ifdef CONFIG_NET_CLS_ACT
322 exts->net = maybe_get_net(exts->net);
323 return exts->net != NULL;
329 static inline void tcf_exts_put_net(struct tcf_exts *exts)
331 #ifdef CONFIG_NET_CLS_ACT
337 #ifdef CONFIG_NET_CLS_ACT
338 #define tcf_exts_for_each_action(i, a, exts) \
339 for (i = 0; i < TCA_ACT_MAX_PRIO && ((a) = (exts)->actions[i]); i++)
341 #define tcf_exts_for_each_action(i, a, exts) \
342 for (; 0; (void)(i), (void)(a), (void)(exts))
346 tcf_exts_stats_update(const struct tcf_exts *exts,
347 u64 bytes, u64 packets, u64 lastuse)
349 #ifdef CONFIG_NET_CLS_ACT
354 for (i = 0; i < exts->nr_actions; i++) {
355 struct tc_action *a = exts->actions[i];
357 tcf_action_stats_update(a, bytes, packets, lastuse, true);
365 * tcf_exts_has_actions - check if at least one action is present
366 * @exts: tc filter extensions handle
368 * Returns true if at least one action is present.
370 static inline bool tcf_exts_has_actions(struct tcf_exts *exts)
372 #ifdef CONFIG_NET_CLS_ACT
373 return exts->nr_actions;
380 * tcf_exts_exec - execute tc filter extensions
381 * @skb: socket buffer
382 * @exts: tc filter extensions handle
383 * @res: desired result
385 * Executes all configured extensions. Returns TC_ACT_OK on a normal execution,
386 * a negative number if the filter must be considered unmatched or
387 * a positive action code (TC_ACT_*) which must be returned to the
391 tcf_exts_exec(struct sk_buff *skb, struct tcf_exts *exts,
392 struct tcf_result *res)
394 #ifdef CONFIG_NET_CLS_ACT
395 return tcf_action_exec(skb, exts->actions, exts->nr_actions, res);
400 int tcf_exts_validate(struct net *net, struct tcf_proto *tp,
401 struct nlattr **tb, struct nlattr *rate_tlv,
402 struct tcf_exts *exts, bool ovr, bool rtnl_held,
403 struct netlink_ext_ack *extack);
404 void tcf_exts_destroy(struct tcf_exts *exts);
405 void tcf_exts_change(struct tcf_exts *dst, struct tcf_exts *src);
406 int tcf_exts_dump(struct sk_buff *skb, struct tcf_exts *exts);
407 int tcf_exts_dump_stats(struct sk_buff *skb, struct tcf_exts *exts);
410 * struct tcf_pkt_info - packet information
412 struct tcf_pkt_info {
417 #ifdef CONFIG_NET_EMATCH
419 struct tcf_ematch_ops;
422 * struct tcf_ematch - extended match (ematch)
424 * @matchid: identifier to allow userspace to reidentify a match
425 * @flags: flags specifying attributes and the relation to other matches
426 * @ops: the operations lookup table of the corresponding ematch module
427 * @datalen: length of the ematch specific configuration data
428 * @data: ematch specific data
431 struct tcf_ematch_ops * ops;
433 unsigned int datalen;
439 static inline int tcf_em_is_container(struct tcf_ematch *em)
444 static inline int tcf_em_is_simple(struct tcf_ematch *em)
446 return em->flags & TCF_EM_SIMPLE;
449 static inline int tcf_em_is_inverted(struct tcf_ematch *em)
451 return em->flags & TCF_EM_INVERT;
454 static inline int tcf_em_last_match(struct tcf_ematch *em)
456 return (em->flags & TCF_EM_REL_MASK) == TCF_EM_REL_END;
459 static inline int tcf_em_early_end(struct tcf_ematch *em, int result)
461 if (tcf_em_last_match(em))
464 if (result == 0 && em->flags & TCF_EM_REL_AND)
467 if (result != 0 && em->flags & TCF_EM_REL_OR)
474 * struct tcf_ematch_tree - ematch tree handle
476 * @hdr: ematch tree header supplied by userspace
477 * @matches: array of ematches
479 struct tcf_ematch_tree {
480 struct tcf_ematch_tree_hdr hdr;
481 struct tcf_ematch * matches;
486 * struct tcf_ematch_ops - ematch module operations
488 * @kind: identifier (kind) of this ematch module
489 * @datalen: length of expected configuration data (optional)
490 * @change: called during validation (optional)
491 * @match: called during ematch tree evaluation, must return 1/0
492 * @destroy: called during destroyage (optional)
493 * @dump: called during dumping process (optional)
494 * @owner: owner, must be set to THIS_MODULE
495 * @link: link to previous/next ematch module (internal use)
497 struct tcf_ematch_ops {
500 int (*change)(struct net *net, void *,
501 int, struct tcf_ematch *);
502 int (*match)(struct sk_buff *, struct tcf_ematch *,
503 struct tcf_pkt_info *);
504 void (*destroy)(struct tcf_ematch *);
505 int (*dump)(struct sk_buff *, struct tcf_ematch *);
506 struct module *owner;
507 struct list_head link;
510 int tcf_em_register(struct tcf_ematch_ops *);
511 void tcf_em_unregister(struct tcf_ematch_ops *);
512 int tcf_em_tree_validate(struct tcf_proto *, struct nlattr *,
513 struct tcf_ematch_tree *);
514 void tcf_em_tree_destroy(struct tcf_ematch_tree *);
515 int tcf_em_tree_dump(struct sk_buff *, struct tcf_ematch_tree *, int);
516 int __tcf_em_tree_match(struct sk_buff *, struct tcf_ematch_tree *,
517 struct tcf_pkt_info *);
520 * tcf_em_tree_match - evaulate an ematch tree
522 * @skb: socket buffer of the packet in question
523 * @tree: ematch tree to be used for evaluation
524 * @info: packet information examined by classifier
526 * This function matches @skb against the ematch tree in @tree by going
527 * through all ematches respecting their logic relations returning
528 * as soon as the result is obvious.
530 * Returns 1 if the ematch tree as-one matches, no ematches are configured
531 * or ematch is not enabled in the kernel, otherwise 0 is returned.
533 static inline int tcf_em_tree_match(struct sk_buff *skb,
534 struct tcf_ematch_tree *tree,
535 struct tcf_pkt_info *info)
537 if (tree->hdr.nmatches)
538 return __tcf_em_tree_match(skb, tree, info);
543 #define MODULE_ALIAS_TCF_EMATCH(kind) MODULE_ALIAS("ematch-kind-" __stringify(kind))
545 #else /* CONFIG_NET_EMATCH */
547 struct tcf_ematch_tree {
550 #define tcf_em_tree_validate(tp, tb, t) ((void)(t), 0)
551 #define tcf_em_tree_destroy(t) do { (void)(t); } while(0)
552 #define tcf_em_tree_dump(skb, t, tlv) (0)
553 #define tcf_em_tree_match(skb, t, info) ((void)(info), 1)
555 #endif /* CONFIG_NET_EMATCH */
557 static inline unsigned char * tcf_get_base_ptr(struct sk_buff *skb, int layer)
561 return skb_mac_header(skb);
562 case TCF_LAYER_NETWORK:
563 return skb_network_header(skb);
564 case TCF_LAYER_TRANSPORT:
565 return skb_transport_header(skb);
571 static inline int tcf_valid_offset(const struct sk_buff *skb,
572 const unsigned char *ptr, const int len)
574 return likely((ptr + len) <= skb_tail_pointer(skb) &&
576 (ptr <= (ptr + len)));
579 #ifdef CONFIG_NET_CLS_IND
580 #include <net/net_namespace.h>
583 tcf_change_indev(struct net *net, struct nlattr *indev_tlv,
584 struct netlink_ext_ack *extack)
586 char indev[IFNAMSIZ];
587 struct net_device *dev;
589 if (nla_strlcpy(indev, indev_tlv, IFNAMSIZ) >= IFNAMSIZ) {
590 NL_SET_ERR_MSG(extack, "Interface name too long");
593 dev = __dev_get_by_name(net, indev);
600 tcf_match_indev(struct sk_buff *skb, int ifindex)
606 return ifindex == skb->skb_iif;
608 #endif /* CONFIG_NET_CLS_IND */
610 int tc_setup_flow_action(struct flow_action *flow_action,
611 const struct tcf_exts *exts);
612 int tc_setup_cb_call(struct tcf_block *block, enum tc_setup_type type,
613 void *type_data, bool err_stop);
614 unsigned int tcf_exts_num_actions(struct tcf_exts *exts);
616 enum tc_block_command {
621 struct tc_block_offload {
622 enum tc_block_command command;
623 enum tcf_block_binder_type binder_type;
624 struct tcf_block *block;
625 struct netlink_ext_ack *extack;
628 struct tc_cls_common_offload {
632 struct netlink_ext_ack *extack;
635 struct tc_cls_u32_knode {
636 struct tcf_exts *exts;
637 struct tcf_result *res;
638 struct tc_u32_sel *sel;
646 struct tc_cls_u32_hnode {
649 unsigned int divisor;
652 enum tc_clsu32_command {
654 TC_CLSU32_REPLACE_KNODE,
655 TC_CLSU32_DELETE_KNODE,
657 TC_CLSU32_REPLACE_HNODE,
658 TC_CLSU32_DELETE_HNODE,
661 struct tc_cls_u32_offload {
662 struct tc_cls_common_offload common;
664 enum tc_clsu32_command command;
666 struct tc_cls_u32_knode knode;
667 struct tc_cls_u32_hnode hnode;
671 static inline bool tc_can_offload(const struct net_device *dev)
673 return dev->features & NETIF_F_HW_TC;
676 static inline bool tc_can_offload_extack(const struct net_device *dev,
677 struct netlink_ext_ack *extack)
679 bool can = tc_can_offload(dev);
682 NL_SET_ERR_MSG(extack, "TC offload is disabled on net device");
688 tc_cls_can_offload_and_chain0(const struct net_device *dev,
689 struct tc_cls_common_offload *common)
691 if (!tc_can_offload_extack(dev, common->extack))
693 if (common->chain_index) {
694 NL_SET_ERR_MSG(common->extack,
695 "Driver supports only offload of chain 0");
701 static inline bool tc_skip_hw(u32 flags)
703 return (flags & TCA_CLS_FLAGS_SKIP_HW) ? true : false;
706 static inline bool tc_skip_sw(u32 flags)
708 return (flags & TCA_CLS_FLAGS_SKIP_SW) ? true : false;
711 /* SKIP_HW and SKIP_SW are mutually exclusive flags. */
712 static inline bool tc_flags_valid(u32 flags)
714 if (flags & ~(TCA_CLS_FLAGS_SKIP_HW | TCA_CLS_FLAGS_SKIP_SW |
715 TCA_CLS_FLAGS_VERBOSE))
718 flags &= TCA_CLS_FLAGS_SKIP_HW | TCA_CLS_FLAGS_SKIP_SW;
719 if (!(flags ^ (TCA_CLS_FLAGS_SKIP_HW | TCA_CLS_FLAGS_SKIP_SW)))
725 static inline bool tc_in_hw(u32 flags)
727 return (flags & TCA_CLS_FLAGS_IN_HW) ? true : false;
731 tc_cls_common_offload_init(struct tc_cls_common_offload *cls_common,
732 const struct tcf_proto *tp, u32 flags,
733 struct netlink_ext_ack *extack)
735 cls_common->chain_index = tp->chain->index;
736 cls_common->protocol = tp->protocol;
737 cls_common->prio = tp->prio;
738 if (tc_skip_sw(flags) || flags & TCA_CLS_FLAGS_VERBOSE)
739 cls_common->extack = extack;
743 TC_CLSFLOWER_REPLACE,
744 TC_CLSFLOWER_DESTROY,
746 TC_CLSFLOWER_TMPLT_CREATE,
747 TC_CLSFLOWER_TMPLT_DESTROY,
750 struct tc_cls_flower_offload {
751 struct tc_cls_common_offload common;
752 enum tc_fl_command command;
753 unsigned long cookie;
754 struct flow_rule *rule;
755 struct flow_stats stats;
759 static inline struct flow_rule *
760 tc_cls_flower_offload_flow_rule(struct tc_cls_flower_offload *tc_flow_cmd)
762 return tc_flow_cmd->rule;
765 enum tc_matchall_command {
766 TC_CLSMATCHALL_REPLACE,
767 TC_CLSMATCHALL_DESTROY,
768 TC_CLSMATCHALL_STATS,
771 struct tc_cls_matchall_offload {
772 struct tc_cls_common_offload common;
773 enum tc_matchall_command command;
774 struct flow_rule *rule;
775 struct flow_stats stats;
776 unsigned long cookie;
779 enum tc_clsbpf_command {
784 struct tc_cls_bpf_offload {
785 struct tc_cls_common_offload common;
786 enum tc_clsbpf_command command;
787 struct tcf_exts *exts;
788 struct bpf_prog *prog;
789 struct bpf_prog *oldprog;
791 bool exts_integrated;
794 struct tc_mqprio_qopt_offload {
795 /* struct tc_mqprio_qopt must always be the first element */
796 struct tc_mqprio_qopt qopt;
800 u64 min_rate[TC_QOPT_MAX_QUEUE];
801 u64 max_rate[TC_QOPT_MAX_QUEUE];
804 /* This structure holds cookie structure that is passed from user
805 * to the kernel for actions and classifiers
813 struct tc_qopt_offload_stats {
814 struct gnet_stats_basic_packed *bstats;
815 struct gnet_stats_queue *qstats;
825 struct tc_mq_opt_offload_graft_params {
830 struct tc_mq_qopt_offload {
831 enum tc_mq_command command;
834 struct tc_qopt_offload_stats stats;
835 struct tc_mq_opt_offload_graft_params graft_params;
839 enum tc_red_command {
847 struct tc_red_qopt_offload_params {
854 struct gnet_stats_queue *qstats;
857 struct tc_red_qopt_offload {
858 enum tc_red_command command;
862 struct tc_red_qopt_offload_params set;
863 struct tc_qopt_offload_stats stats;
864 struct red_stats *xstats;
869 enum tc_gred_command {
875 struct tc_gred_vq_qopt_offload_params {
884 /* Only need backlog, see struct tc_prio_qopt_offload_params */
888 struct tc_gred_qopt_offload_params {
893 struct gnet_stats_queue *qstats;
894 struct tc_gred_vq_qopt_offload_params tab[MAX_DPs];
897 struct tc_gred_qopt_offload_stats {
898 struct gnet_stats_basic_packed bstats[MAX_DPs];
899 struct gnet_stats_queue qstats[MAX_DPs];
900 struct red_stats *xstats[MAX_DPs];
903 struct tc_gred_qopt_offload {
904 enum tc_gred_command command;
908 struct tc_gred_qopt_offload_params set;
909 struct tc_gred_qopt_offload_stats stats;
913 enum tc_prio_command {
920 struct tc_prio_qopt_offload_params {
922 u8 priomap[TC_PRIO_MAX + 1];
923 /* In case that a prio qdisc is offloaded and now is changed to a
924 * non-offloadedable config, it needs to update the backlog & qlen
925 * values to negate the HW backlog & qlen values (and only them).
927 struct gnet_stats_queue *qstats;
930 struct tc_prio_qopt_offload_graft_params {
935 struct tc_prio_qopt_offload {
936 enum tc_prio_command command;
940 struct tc_prio_qopt_offload_params replace_params;
941 struct tc_qopt_offload_stats stats;
942 struct tc_prio_qopt_offload_graft_params graft_params;
946 enum tc_root_command {
950 struct tc_root_qopt_offload {
951 enum tc_root_command command;