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 __tc_indr_block_cb_register(struct net_device *dev, void *cb_priv,
74 tc_indr_block_bind_cb_t *cb, void *cb_ident);
75 int tc_indr_block_cb_register(struct net_device *dev, void *cb_priv,
76 tc_indr_block_bind_cb_t *cb, void *cb_ident);
77 void __tc_indr_block_cb_unregister(struct net_device *dev,
78 tc_indr_block_bind_cb_t *cb, void *cb_ident);
79 void tc_indr_block_cb_unregister(struct net_device *dev,
80 tc_indr_block_bind_cb_t *cb, void *cb_ident);
82 int tcf_classify(struct sk_buff *skb, const struct tcf_proto *tp,
83 struct tcf_result *res, bool compat_mode);
86 static inline bool tcf_block_shared(struct tcf_block *block)
91 static inline bool tcf_block_non_null_shared(struct tcf_block *block)
97 int tcf_block_get(struct tcf_block **p_block,
98 struct tcf_proto __rcu **p_filter_chain, struct Qdisc *q,
99 struct netlink_ext_ack *extack)
105 int tcf_block_get_ext(struct tcf_block **p_block, struct Qdisc *q,
106 struct tcf_block_ext_info *ei,
107 struct netlink_ext_ack *extack)
112 static inline void tcf_block_put(struct tcf_block *block)
117 void tcf_block_put_ext(struct tcf_block *block, struct Qdisc *q,
118 struct tcf_block_ext_info *ei)
122 static inline struct Qdisc *tcf_block_q(struct tcf_block *block)
128 int tc_setup_cb_block_register(struct tcf_block *block, flow_setup_cb_t *cb,
135 void tc_setup_cb_block_unregister(struct tcf_block *block, flow_setup_cb_t *cb,
141 int __tc_indr_block_cb_register(struct net_device *dev, void *cb_priv,
142 tc_indr_block_bind_cb_t *cb, void *cb_ident)
148 int tc_indr_block_cb_register(struct net_device *dev, void *cb_priv,
149 tc_indr_block_bind_cb_t *cb, void *cb_ident)
155 void __tc_indr_block_cb_unregister(struct net_device *dev,
156 tc_indr_block_bind_cb_t *cb, void *cb_ident)
161 void tc_indr_block_cb_unregister(struct net_device *dev,
162 tc_indr_block_bind_cb_t *cb, void *cb_ident)
166 static inline int tcf_classify(struct sk_buff *skb, const struct tcf_proto *tp,
167 struct tcf_result *res, bool compat_mode)
169 return TC_ACT_UNSPEC;
173 static inline unsigned long
174 __cls_set_class(unsigned long *clp, unsigned long cl)
176 return xchg(clp, cl);
179 static inline unsigned long
180 cls_set_class(struct Qdisc *q, unsigned long *clp, unsigned long cl)
182 unsigned long old_cl;
185 old_cl = __cls_set_class(clp, cl);
191 tcf_bind_filter(struct tcf_proto *tp, struct tcf_result *r, unsigned long base)
193 struct Qdisc *q = tp->chain->block->q;
196 /* Check q as it is not set for shared blocks. In that case,
197 * setting class is not supported.
201 cl = q->ops->cl_ops->bind_tcf(q, base, r->classid);
202 cl = cls_set_class(q, &r->class, cl);
204 q->ops->cl_ops->unbind_tcf(q, cl);
208 tcf_unbind_filter(struct tcf_proto *tp, struct tcf_result *r)
210 struct Qdisc *q = tp->chain->block->q;
215 if ((cl = __cls_set_class(&r->class, 0)) != 0)
216 q->ops->cl_ops->unbind_tcf(q, cl);
220 #ifdef CONFIG_NET_CLS_ACT
221 __u32 type; /* for backward compat(TCA_OLD_COMPAT) */
223 struct tc_action **actions;
226 /* Map to export classifier specific extension TLV types to the
227 * generic extensions API. Unsupported extensions must be set to 0.
233 static inline int tcf_exts_init(struct tcf_exts *exts, struct net *net,
234 int action, int police)
236 #ifdef CONFIG_NET_CLS_ACT
238 exts->nr_actions = 0;
240 exts->actions = kcalloc(TCA_ACT_MAX_PRIO, sizeof(struct tc_action *),
245 exts->action = action;
246 exts->police = police;
250 /* Return false if the netns is being destroyed in cleanup_net(). Callers
251 * need to do cleanup synchronously in this case, otherwise may race with
252 * tc_action_net_exit(). Return true for other cases.
254 static inline bool tcf_exts_get_net(struct tcf_exts *exts)
256 #ifdef CONFIG_NET_CLS_ACT
257 exts->net = maybe_get_net(exts->net);
258 return exts->net != NULL;
264 static inline void tcf_exts_put_net(struct tcf_exts *exts)
266 #ifdef CONFIG_NET_CLS_ACT
272 #ifdef CONFIG_NET_CLS_ACT
273 #define tcf_exts_for_each_action(i, a, exts) \
274 for (i = 0; i < TCA_ACT_MAX_PRIO && ((a) = (exts)->actions[i]); i++)
276 #define tcf_exts_for_each_action(i, a, exts) \
277 for (; 0; (void)(i), (void)(a), (void)(exts))
281 tcf_exts_stats_update(const struct tcf_exts *exts,
282 u64 bytes, u64 packets, u64 lastuse)
284 #ifdef CONFIG_NET_CLS_ACT
289 for (i = 0; i < exts->nr_actions; i++) {
290 struct tc_action *a = exts->actions[i];
292 tcf_action_stats_update(a, bytes, packets, lastuse, true);
300 * tcf_exts_has_actions - check if at least one action is present
301 * @exts: tc filter extensions handle
303 * Returns true if at least one action is present.
305 static inline bool tcf_exts_has_actions(struct tcf_exts *exts)
307 #ifdef CONFIG_NET_CLS_ACT
308 return exts->nr_actions;
315 * tcf_exts_exec - execute tc filter extensions
316 * @skb: socket buffer
317 * @exts: tc filter extensions handle
318 * @res: desired result
320 * Executes all configured extensions. Returns TC_ACT_OK on a normal execution,
321 * a negative number if the filter must be considered unmatched or
322 * a positive action code (TC_ACT_*) which must be returned to the
326 tcf_exts_exec(struct sk_buff *skb, struct tcf_exts *exts,
327 struct tcf_result *res)
329 #ifdef CONFIG_NET_CLS_ACT
330 return tcf_action_exec(skb, exts->actions, exts->nr_actions, res);
335 int tcf_exts_validate(struct net *net, struct tcf_proto *tp,
336 struct nlattr **tb, struct nlattr *rate_tlv,
337 struct tcf_exts *exts, bool ovr, bool rtnl_held,
338 struct netlink_ext_ack *extack);
339 void tcf_exts_destroy(struct tcf_exts *exts);
340 void tcf_exts_change(struct tcf_exts *dst, struct tcf_exts *src);
341 int tcf_exts_dump(struct sk_buff *skb, struct tcf_exts *exts);
342 int tcf_exts_dump_stats(struct sk_buff *skb, struct tcf_exts *exts);
345 * struct tcf_pkt_info - packet information
347 struct tcf_pkt_info {
352 #ifdef CONFIG_NET_EMATCH
354 struct tcf_ematch_ops;
357 * struct tcf_ematch - extended match (ematch)
359 * @matchid: identifier to allow userspace to reidentify a match
360 * @flags: flags specifying attributes and the relation to other matches
361 * @ops: the operations lookup table of the corresponding ematch module
362 * @datalen: length of the ematch specific configuration data
363 * @data: ematch specific data
366 struct tcf_ematch_ops * ops;
368 unsigned int datalen;
374 static inline int tcf_em_is_container(struct tcf_ematch *em)
379 static inline int tcf_em_is_simple(struct tcf_ematch *em)
381 return em->flags & TCF_EM_SIMPLE;
384 static inline int tcf_em_is_inverted(struct tcf_ematch *em)
386 return em->flags & TCF_EM_INVERT;
389 static inline int tcf_em_last_match(struct tcf_ematch *em)
391 return (em->flags & TCF_EM_REL_MASK) == TCF_EM_REL_END;
394 static inline int tcf_em_early_end(struct tcf_ematch *em, int result)
396 if (tcf_em_last_match(em))
399 if (result == 0 && em->flags & TCF_EM_REL_AND)
402 if (result != 0 && em->flags & TCF_EM_REL_OR)
409 * struct tcf_ematch_tree - ematch tree handle
411 * @hdr: ematch tree header supplied by userspace
412 * @matches: array of ematches
414 struct tcf_ematch_tree {
415 struct tcf_ematch_tree_hdr hdr;
416 struct tcf_ematch * matches;
421 * struct tcf_ematch_ops - ematch module operations
423 * @kind: identifier (kind) of this ematch module
424 * @datalen: length of expected configuration data (optional)
425 * @change: called during validation (optional)
426 * @match: called during ematch tree evaluation, must return 1/0
427 * @destroy: called during destroyage (optional)
428 * @dump: called during dumping process (optional)
429 * @owner: owner, must be set to THIS_MODULE
430 * @link: link to previous/next ematch module (internal use)
432 struct tcf_ematch_ops {
435 int (*change)(struct net *net, void *,
436 int, struct tcf_ematch *);
437 int (*match)(struct sk_buff *, struct tcf_ematch *,
438 struct tcf_pkt_info *);
439 void (*destroy)(struct tcf_ematch *);
440 int (*dump)(struct sk_buff *, struct tcf_ematch *);
441 struct module *owner;
442 struct list_head link;
445 int tcf_em_register(struct tcf_ematch_ops *);
446 void tcf_em_unregister(struct tcf_ematch_ops *);
447 int tcf_em_tree_validate(struct tcf_proto *, struct nlattr *,
448 struct tcf_ematch_tree *);
449 void tcf_em_tree_destroy(struct tcf_ematch_tree *);
450 int tcf_em_tree_dump(struct sk_buff *, struct tcf_ematch_tree *, int);
451 int __tcf_em_tree_match(struct sk_buff *, struct tcf_ematch_tree *,
452 struct tcf_pkt_info *);
455 * tcf_em_tree_match - evaulate an ematch tree
457 * @skb: socket buffer of the packet in question
458 * @tree: ematch tree to be used for evaluation
459 * @info: packet information examined by classifier
461 * This function matches @skb against the ematch tree in @tree by going
462 * through all ematches respecting their logic relations returning
463 * as soon as the result is obvious.
465 * Returns 1 if the ematch tree as-one matches, no ematches are configured
466 * or ematch is not enabled in the kernel, otherwise 0 is returned.
468 static inline int tcf_em_tree_match(struct sk_buff *skb,
469 struct tcf_ematch_tree *tree,
470 struct tcf_pkt_info *info)
472 if (tree->hdr.nmatches)
473 return __tcf_em_tree_match(skb, tree, info);
478 #define MODULE_ALIAS_TCF_EMATCH(kind) MODULE_ALIAS("ematch-kind-" __stringify(kind))
480 #else /* CONFIG_NET_EMATCH */
482 struct tcf_ematch_tree {
485 #define tcf_em_tree_validate(tp, tb, t) ((void)(t), 0)
486 #define tcf_em_tree_destroy(t) do { (void)(t); } while(0)
487 #define tcf_em_tree_dump(skb, t, tlv) (0)
488 #define tcf_em_tree_match(skb, t, info) ((void)(info), 1)
490 #endif /* CONFIG_NET_EMATCH */
492 static inline unsigned char * tcf_get_base_ptr(struct sk_buff *skb, int layer)
496 return skb_mac_header(skb);
497 case TCF_LAYER_NETWORK:
498 return skb_network_header(skb);
499 case TCF_LAYER_TRANSPORT:
500 return skb_transport_header(skb);
506 static inline int tcf_valid_offset(const struct sk_buff *skb,
507 const unsigned char *ptr, const int len)
509 return likely((ptr + len) <= skb_tail_pointer(skb) &&
511 (ptr <= (ptr + len)));
515 tcf_change_indev(struct net *net, struct nlattr *indev_tlv,
516 struct netlink_ext_ack *extack)
518 char indev[IFNAMSIZ];
519 struct net_device *dev;
521 if (nla_strlcpy(indev, indev_tlv, IFNAMSIZ) >= IFNAMSIZ) {
522 NL_SET_ERR_MSG(extack, "Interface name too long");
525 dev = __dev_get_by_name(net, indev);
532 tcf_match_indev(struct sk_buff *skb, int ifindex)
538 return ifindex == skb->skb_iif;
541 int tc_setup_flow_action(struct flow_action *flow_action,
542 const struct tcf_exts *exts);
543 int tc_setup_cb_call(struct tcf_block *block, enum tc_setup_type type,
544 void *type_data, bool err_stop);
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;
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 /* In case that a prio qdisc is offloaded and now is changed to a
813 * non-offloadedable config, it needs to update the backlog & qlen
814 * values to negate the HW backlog & qlen values (and only them).
816 struct gnet_stats_queue *qstats;
819 struct tc_prio_qopt_offload_graft_params {
824 struct tc_prio_qopt_offload {
825 enum tc_prio_command command;
829 struct tc_prio_qopt_offload_params replace_params;
830 struct tc_qopt_offload_stats stats;
831 struct tc_prio_qopt_offload_graft_params graft_params;
835 enum tc_root_command {
839 struct tc_root_qopt_offload {
840 enum tc_root_command command;