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)
267 #ifdef CONFIG_NET_CLS_ACT
272 for (i = 0; i < exts->nr_actions; i++) {
273 struct tc_action *a = exts->actions[i];
275 tcf_action_stats_update(a, bytes, packets, lastuse, true);
283 * tcf_exts_has_actions - check if at least one action is present
284 * @exts: tc filter extensions handle
286 * Returns true if at least one action is present.
288 static inline bool tcf_exts_has_actions(struct tcf_exts *exts)
290 #ifdef CONFIG_NET_CLS_ACT
291 return exts->nr_actions;
298 * tcf_exts_exec - execute tc filter extensions
299 * @skb: socket buffer
300 * @exts: tc filter extensions handle
301 * @res: desired result
303 * Executes all configured extensions. Returns TC_ACT_OK on a normal execution,
304 * a negative number if the filter must be considered unmatched or
305 * a positive action code (TC_ACT_*) which must be returned to the
309 tcf_exts_exec(struct sk_buff *skb, struct tcf_exts *exts,
310 struct tcf_result *res)
312 #ifdef CONFIG_NET_CLS_ACT
313 return tcf_action_exec(skb, exts->actions, exts->nr_actions, res);
318 int tcf_exts_validate(struct net *net, struct tcf_proto *tp,
319 struct nlattr **tb, struct nlattr *rate_tlv,
320 struct tcf_exts *exts, bool ovr, bool rtnl_held,
321 struct netlink_ext_ack *extack);
322 void tcf_exts_destroy(struct tcf_exts *exts);
323 void tcf_exts_change(struct tcf_exts *dst, struct tcf_exts *src);
324 int tcf_exts_dump(struct sk_buff *skb, struct tcf_exts *exts);
325 int tcf_exts_dump_stats(struct sk_buff *skb, struct tcf_exts *exts);
328 * struct tcf_pkt_info - packet information
330 struct tcf_pkt_info {
335 #ifdef CONFIG_NET_EMATCH
337 struct tcf_ematch_ops;
340 * struct tcf_ematch - extended match (ematch)
342 * @matchid: identifier to allow userspace to reidentify a match
343 * @flags: flags specifying attributes and the relation to other matches
344 * @ops: the operations lookup table of the corresponding ematch module
345 * @datalen: length of the ematch specific configuration data
346 * @data: ematch specific data
349 struct tcf_ematch_ops * ops;
351 unsigned int datalen;
357 static inline int tcf_em_is_container(struct tcf_ematch *em)
362 static inline int tcf_em_is_simple(struct tcf_ematch *em)
364 return em->flags & TCF_EM_SIMPLE;
367 static inline int tcf_em_is_inverted(struct tcf_ematch *em)
369 return em->flags & TCF_EM_INVERT;
372 static inline int tcf_em_last_match(struct tcf_ematch *em)
374 return (em->flags & TCF_EM_REL_MASK) == TCF_EM_REL_END;
377 static inline int tcf_em_early_end(struct tcf_ematch *em, int result)
379 if (tcf_em_last_match(em))
382 if (result == 0 && em->flags & TCF_EM_REL_AND)
385 if (result != 0 && em->flags & TCF_EM_REL_OR)
392 * struct tcf_ematch_tree - ematch tree handle
394 * @hdr: ematch tree header supplied by userspace
395 * @matches: array of ematches
397 struct tcf_ematch_tree {
398 struct tcf_ematch_tree_hdr hdr;
399 struct tcf_ematch * matches;
404 * struct tcf_ematch_ops - ematch module operations
406 * @kind: identifier (kind) of this ematch module
407 * @datalen: length of expected configuration data (optional)
408 * @change: called during validation (optional)
409 * @match: called during ematch tree evaluation, must return 1/0
410 * @destroy: called during destroyage (optional)
411 * @dump: called during dumping process (optional)
412 * @owner: owner, must be set to THIS_MODULE
413 * @link: link to previous/next ematch module (internal use)
415 struct tcf_ematch_ops {
418 int (*change)(struct net *net, void *,
419 int, struct tcf_ematch *);
420 int (*match)(struct sk_buff *, struct tcf_ematch *,
421 struct tcf_pkt_info *);
422 void (*destroy)(struct tcf_ematch *);
423 int (*dump)(struct sk_buff *, struct tcf_ematch *);
424 struct module *owner;
425 struct list_head link;
428 int tcf_em_register(struct tcf_ematch_ops *);
429 void tcf_em_unregister(struct tcf_ematch_ops *);
430 int tcf_em_tree_validate(struct tcf_proto *, struct nlattr *,
431 struct tcf_ematch_tree *);
432 void tcf_em_tree_destroy(struct tcf_ematch_tree *);
433 int tcf_em_tree_dump(struct sk_buff *, struct tcf_ematch_tree *, int);
434 int __tcf_em_tree_match(struct sk_buff *, struct tcf_ematch_tree *,
435 struct tcf_pkt_info *);
438 * tcf_em_tree_match - evaulate an ematch tree
440 * @skb: socket buffer of the packet in question
441 * @tree: ematch tree to be used for evaluation
442 * @info: packet information examined by classifier
444 * This function matches @skb against the ematch tree in @tree by going
445 * through all ematches respecting their logic relations returning
446 * as soon as the result is obvious.
448 * Returns 1 if the ematch tree as-one matches, no ematches are configured
449 * or ematch is not enabled in the kernel, otherwise 0 is returned.
451 static inline int tcf_em_tree_match(struct sk_buff *skb,
452 struct tcf_ematch_tree *tree,
453 struct tcf_pkt_info *info)
455 if (tree->hdr.nmatches)
456 return __tcf_em_tree_match(skb, tree, info);
461 #define MODULE_ALIAS_TCF_EMATCH(kind) MODULE_ALIAS("ematch-kind-" __stringify(kind))
463 #else /* CONFIG_NET_EMATCH */
465 struct tcf_ematch_tree {
468 #define tcf_em_tree_validate(tp, tb, t) ((void)(t), 0)
469 #define tcf_em_tree_destroy(t) do { (void)(t); } while(0)
470 #define tcf_em_tree_dump(skb, t, tlv) (0)
471 #define tcf_em_tree_match(skb, t, info) ((void)(info), 1)
473 #endif /* CONFIG_NET_EMATCH */
475 static inline unsigned char * tcf_get_base_ptr(struct sk_buff *skb, int layer)
479 return skb_mac_header(skb);
480 case TCF_LAYER_NETWORK:
481 return skb_network_header(skb);
482 case TCF_LAYER_TRANSPORT:
483 return skb_transport_header(skb);
489 static inline int tcf_valid_offset(const struct sk_buff *skb,
490 const unsigned char *ptr, const int len)
492 return likely((ptr + len) <= skb_tail_pointer(skb) &&
494 (ptr <= (ptr + len)));
498 tcf_change_indev(struct net *net, struct nlattr *indev_tlv,
499 struct netlink_ext_ack *extack)
501 char indev[IFNAMSIZ];
502 struct net_device *dev;
504 if (nla_strlcpy(indev, indev_tlv, IFNAMSIZ) >= IFNAMSIZ) {
505 NL_SET_ERR_MSG(extack, "Interface name too long");
508 dev = __dev_get_by_name(net, indev);
515 tcf_match_indev(struct sk_buff *skb, int ifindex)
521 return ifindex == skb->skb_iif;
524 int tc_setup_flow_action(struct flow_action *flow_action,
525 const struct tcf_exts *exts);
526 void tc_cleanup_flow_action(struct flow_action *flow_action);
528 int tc_setup_cb_call(struct tcf_block *block, enum tc_setup_type type,
529 void *type_data, bool err_stop, bool rtnl_held);
530 int tc_setup_cb_add(struct tcf_block *block, struct tcf_proto *tp,
531 enum tc_setup_type type, void *type_data, bool err_stop,
532 u32 *flags, unsigned int *in_hw_count, bool rtnl_held);
533 int tc_setup_cb_replace(struct tcf_block *block, struct tcf_proto *tp,
534 enum tc_setup_type type, void *type_data, bool err_stop,
535 u32 *old_flags, unsigned int *old_in_hw_count,
536 u32 *new_flags, unsigned int *new_in_hw_count,
538 int tc_setup_cb_destroy(struct tcf_block *block, struct tcf_proto *tp,
539 enum tc_setup_type type, void *type_data, bool err_stop,
540 u32 *flags, unsigned int *in_hw_count, bool rtnl_held);
541 int tc_setup_cb_reoffload(struct tcf_block *block, struct tcf_proto *tp,
542 bool add, flow_setup_cb_t *cb,
543 enum tc_setup_type type, void *type_data,
544 void *cb_priv, u32 *flags, unsigned int *in_hw_count);
545 unsigned int tcf_exts_num_actions(struct tcf_exts *exts);
547 struct tc_cls_u32_knode {
548 struct tcf_exts *exts;
549 struct tcf_result *res;
550 struct tc_u32_sel *sel;
558 struct tc_cls_u32_hnode {
561 unsigned int divisor;
564 enum tc_clsu32_command {
566 TC_CLSU32_REPLACE_KNODE,
567 TC_CLSU32_DELETE_KNODE,
569 TC_CLSU32_REPLACE_HNODE,
570 TC_CLSU32_DELETE_HNODE,
573 struct tc_cls_u32_offload {
574 struct flow_cls_common_offload common;
576 enum tc_clsu32_command command;
578 struct tc_cls_u32_knode knode;
579 struct tc_cls_u32_hnode hnode;
583 static inline bool tc_can_offload(const struct net_device *dev)
585 return dev->features & NETIF_F_HW_TC;
588 static inline bool tc_can_offload_extack(const struct net_device *dev,
589 struct netlink_ext_ack *extack)
591 bool can = tc_can_offload(dev);
594 NL_SET_ERR_MSG(extack, "TC offload is disabled on net device");
600 tc_cls_can_offload_and_chain0(const struct net_device *dev,
601 struct flow_cls_common_offload *common)
603 if (!tc_can_offload_extack(dev, common->extack))
605 if (common->chain_index) {
606 NL_SET_ERR_MSG(common->extack,
607 "Driver supports only offload of chain 0");
613 static inline bool tc_skip_hw(u32 flags)
615 return (flags & TCA_CLS_FLAGS_SKIP_HW) ? true : false;
618 static inline bool tc_skip_sw(u32 flags)
620 return (flags & TCA_CLS_FLAGS_SKIP_SW) ? true : false;
623 /* SKIP_HW and SKIP_SW are mutually exclusive flags. */
624 static inline bool tc_flags_valid(u32 flags)
626 if (flags & ~(TCA_CLS_FLAGS_SKIP_HW | TCA_CLS_FLAGS_SKIP_SW |
627 TCA_CLS_FLAGS_VERBOSE))
630 flags &= TCA_CLS_FLAGS_SKIP_HW | TCA_CLS_FLAGS_SKIP_SW;
631 if (!(flags ^ (TCA_CLS_FLAGS_SKIP_HW | TCA_CLS_FLAGS_SKIP_SW)))
637 static inline bool tc_in_hw(u32 flags)
639 return (flags & TCA_CLS_FLAGS_IN_HW) ? true : false;
643 tc_cls_common_offload_init(struct flow_cls_common_offload *cls_common,
644 const struct tcf_proto *tp, u32 flags,
645 struct netlink_ext_ack *extack)
647 cls_common->chain_index = tp->chain->index;
648 cls_common->protocol = tp->protocol;
649 cls_common->prio = tp->prio >> 16;
650 if (tc_skip_sw(flags) || flags & TCA_CLS_FLAGS_VERBOSE)
651 cls_common->extack = extack;
654 enum tc_matchall_command {
655 TC_CLSMATCHALL_REPLACE,
656 TC_CLSMATCHALL_DESTROY,
657 TC_CLSMATCHALL_STATS,
660 struct tc_cls_matchall_offload {
661 struct flow_cls_common_offload common;
662 enum tc_matchall_command command;
663 struct flow_rule *rule;
664 struct flow_stats stats;
665 unsigned long cookie;
668 enum tc_clsbpf_command {
673 struct tc_cls_bpf_offload {
674 struct flow_cls_common_offload common;
675 enum tc_clsbpf_command command;
676 struct tcf_exts *exts;
677 struct bpf_prog *prog;
678 struct bpf_prog *oldprog;
680 bool exts_integrated;
683 struct tc_mqprio_qopt_offload {
684 /* struct tc_mqprio_qopt must always be the first element */
685 struct tc_mqprio_qopt qopt;
689 u64 min_rate[TC_QOPT_MAX_QUEUE];
690 u64 max_rate[TC_QOPT_MAX_QUEUE];
693 /* This structure holds cookie structure that is passed from user
694 * to the kernel for actions and classifiers
702 struct tc_qopt_offload_stats {
703 struct gnet_stats_basic_packed *bstats;
704 struct gnet_stats_queue *qstats;
714 struct tc_mq_opt_offload_graft_params {
719 struct tc_mq_qopt_offload {
720 enum tc_mq_command command;
723 struct tc_qopt_offload_stats stats;
724 struct tc_mq_opt_offload_graft_params graft_params;
728 enum tc_red_command {
736 struct tc_red_qopt_offload_params {
743 struct gnet_stats_queue *qstats;
746 struct tc_red_qopt_offload {
747 enum tc_red_command command;
751 struct tc_red_qopt_offload_params set;
752 struct tc_qopt_offload_stats stats;
753 struct red_stats *xstats;
758 enum tc_gred_command {
764 struct tc_gred_vq_qopt_offload_params {
773 /* Only need backlog, see struct tc_prio_qopt_offload_params */
777 struct tc_gred_qopt_offload_params {
782 struct gnet_stats_queue *qstats;
783 struct tc_gred_vq_qopt_offload_params tab[MAX_DPs];
786 struct tc_gred_qopt_offload_stats {
787 struct gnet_stats_basic_packed bstats[MAX_DPs];
788 struct gnet_stats_queue qstats[MAX_DPs];
789 struct red_stats *xstats[MAX_DPs];
792 struct tc_gred_qopt_offload {
793 enum tc_gred_command command;
797 struct tc_gred_qopt_offload_params set;
798 struct tc_gred_qopt_offload_stats stats;
802 enum tc_prio_command {
809 struct tc_prio_qopt_offload_params {
811 u8 priomap[TC_PRIO_MAX + 1];
812 /* At the point of un-offloading the Qdisc, the reported backlog and
813 * qlen need to be reduced by the portion that is in HW.
815 struct gnet_stats_queue *qstats;
818 struct tc_prio_qopt_offload_graft_params {
823 struct tc_prio_qopt_offload {
824 enum tc_prio_command command;
828 struct tc_prio_qopt_offload_params replace_params;
829 struct tc_qopt_offload_stats stats;
830 struct tc_prio_qopt_offload_graft_params graft_params;
834 enum tc_root_command {
838 struct tc_root_qopt_offload {
839 enum tc_root_command command;
844 enum tc_ets_command {
851 struct tc_ets_qopt_offload_replace_params {
853 u8 priomap[TC_PRIO_MAX + 1];
854 unsigned int quanta[TCQ_ETS_MAX_BANDS]; /* 0 for strict bands. */
855 unsigned int weights[TCQ_ETS_MAX_BANDS];
856 struct gnet_stats_queue *qstats;
859 struct tc_ets_qopt_offload_graft_params {
864 struct tc_ets_qopt_offload {
865 enum tc_ets_command command;
869 struct tc_ets_qopt_offload_replace_params replace_params;
870 struct tc_qopt_offload_stats stats;
871 struct tc_ets_qopt_offload_graft_params graft_params;
875 enum tc_tbf_command {
881 struct tc_tbf_qopt_offload_replace_params {
882 struct psched_ratecfg rate;
884 struct gnet_stats_queue *qstats;
887 struct tc_tbf_qopt_offload {
888 enum tc_tbf_command command;
892 struct tc_tbf_qopt_offload_replace_params replace_params;
893 struct tc_qopt_offload_stats stats;
897 enum tc_fifo_command {
903 struct tc_fifo_qopt_offload {
904 enum tc_fifo_command command;
908 struct tc_qopt_offload_stats stats;