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>
10 #include <net/net_namespace.h>
12 /* TC action not accessible from user space */
13 #define TC_ACT_CONSUMED (TC_ACT_VALUE_MAX + 1)
15 /* Basic packet classifier frontend definitions. */
23 int (*fn)(struct tcf_proto *, void *node, struct tcf_walker *);
26 int register_tcf_proto_ops(struct tcf_proto_ops *ops);
27 int unregister_tcf_proto_ops(struct tcf_proto_ops *ops);
29 struct tcf_block_ext_info {
30 enum flow_block_binder_type binder_type;
31 tcf_chain_head_change_t *chain_head_change;
32 void *chain_head_change_priv;
37 bool tcf_queue_work(struct rcu_work *rwork, work_func_t func);
40 struct tcf_chain *tcf_chain_get_by_act(struct tcf_block *block,
42 void tcf_chain_put_by_act(struct tcf_chain *chain);
43 struct tcf_chain *tcf_get_next_chain(struct tcf_block *block,
44 struct tcf_chain *chain);
45 struct tcf_proto *tcf_get_next_proto(struct tcf_chain *chain,
46 struct tcf_proto *tp, bool rtnl_held);
47 void tcf_block_netif_keep_dst(struct tcf_block *block);
48 int tcf_block_get(struct tcf_block **p_block,
49 struct tcf_proto __rcu **p_filter_chain, struct Qdisc *q,
50 struct netlink_ext_ack *extack);
51 int tcf_block_get_ext(struct tcf_block **p_block, struct Qdisc *q,
52 struct tcf_block_ext_info *ei,
53 struct netlink_ext_ack *extack);
54 void tcf_block_put(struct tcf_block *block);
55 void tcf_block_put_ext(struct tcf_block *block, struct Qdisc *q,
56 struct tcf_block_ext_info *ei);
58 static inline bool tcf_block_shared(struct tcf_block *block)
63 static inline struct Qdisc *tcf_block_q(struct tcf_block *block)
65 WARN_ON(tcf_block_shared(block));
69 int __tc_indr_block_cb_register(struct net_device *dev, void *cb_priv,
70 tc_indr_block_bind_cb_t *cb, void *cb_ident);
71 int tc_indr_block_cb_register(struct net_device *dev, void *cb_priv,
72 tc_indr_block_bind_cb_t *cb, void *cb_ident);
73 void __tc_indr_block_cb_unregister(struct net_device *dev,
74 tc_indr_block_bind_cb_t *cb, void *cb_ident);
75 void tc_indr_block_cb_unregister(struct net_device *dev,
76 tc_indr_block_bind_cb_t *cb, void *cb_ident);
78 int tcf_classify(struct sk_buff *skb, const struct tcf_proto *tp,
79 struct tcf_result *res, bool compat_mode);
82 static inline bool tcf_block_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, tc_setup_cb_t *cb,
126 void tc_setup_cb_block_unregister(struct tcf_block *block, tc_setup_cb_t *cb,
132 int __tc_indr_block_cb_register(struct net_device *dev, void *cb_priv,
133 tc_indr_block_bind_cb_t *cb, void *cb_ident)
139 int tc_indr_block_cb_register(struct net_device *dev, void *cb_priv,
140 tc_indr_block_bind_cb_t *cb, void *cb_ident)
146 void __tc_indr_block_cb_unregister(struct net_device *dev,
147 tc_indr_block_bind_cb_t *cb, void *cb_ident)
152 void tc_indr_block_cb_unregister(struct net_device *dev,
153 tc_indr_block_bind_cb_t *cb, void *cb_ident)
157 static inline int tcf_classify(struct sk_buff *skb, const struct tcf_proto *tp,
158 struct tcf_result *res, bool compat_mode)
160 return TC_ACT_UNSPEC;
164 static inline unsigned long
165 __cls_set_class(unsigned long *clp, unsigned long cl)
167 return xchg(clp, cl);
170 static inline unsigned long
171 cls_set_class(struct Qdisc *q, unsigned long *clp, unsigned long cl)
173 unsigned long old_cl;
176 old_cl = __cls_set_class(clp, cl);
182 tcf_bind_filter(struct tcf_proto *tp, struct tcf_result *r, unsigned long base)
184 struct Qdisc *q = tp->chain->block->q;
187 /* Check q as it is not set for shared blocks. In that case,
188 * setting class is not supported.
192 cl = q->ops->cl_ops->bind_tcf(q, base, r->classid);
193 cl = cls_set_class(q, &r->class, cl);
195 q->ops->cl_ops->unbind_tcf(q, cl);
199 tcf_unbind_filter(struct tcf_proto *tp, struct tcf_result *r)
201 struct Qdisc *q = tp->chain->block->q;
206 if ((cl = __cls_set_class(&r->class, 0)) != 0)
207 q->ops->cl_ops->unbind_tcf(q, cl);
211 #ifdef CONFIG_NET_CLS_ACT
212 __u32 type; /* for backward compat(TCA_OLD_COMPAT) */
214 struct tc_action **actions;
217 /* Map to export classifier specific extension TLV types to the
218 * generic extensions API. Unsupported extensions must be set to 0.
224 static inline int tcf_exts_init(struct tcf_exts *exts, struct net *net,
225 int action, int police)
227 #ifdef CONFIG_NET_CLS_ACT
229 exts->nr_actions = 0;
231 exts->actions = kcalloc(TCA_ACT_MAX_PRIO, sizeof(struct tc_action *),
236 exts->action = action;
237 exts->police = police;
241 /* Return false if the netns is being destroyed in cleanup_net(). Callers
242 * need to do cleanup synchronously in this case, otherwise may race with
243 * tc_action_net_exit(). Return true for other cases.
245 static inline bool tcf_exts_get_net(struct tcf_exts *exts)
247 #ifdef CONFIG_NET_CLS_ACT
248 exts->net = maybe_get_net(exts->net);
249 return exts->net != NULL;
255 static inline void tcf_exts_put_net(struct tcf_exts *exts)
257 #ifdef CONFIG_NET_CLS_ACT
263 #ifdef CONFIG_NET_CLS_ACT
264 #define tcf_exts_for_each_action(i, a, exts) \
265 for (i = 0; i < TCA_ACT_MAX_PRIO && ((a) = (exts)->actions[i]); i++)
267 #define tcf_exts_for_each_action(i, a, exts) \
268 for (; 0; (void)(i), (void)(a), (void)(exts))
272 tcf_exts_stats_update(const struct tcf_exts *exts,
273 u64 bytes, u64 packets, u64 lastuse)
275 #ifdef CONFIG_NET_CLS_ACT
280 for (i = 0; i < exts->nr_actions; i++) {
281 struct tc_action *a = exts->actions[i];
283 tcf_action_stats_update(a, bytes, packets, lastuse, true);
291 * tcf_exts_has_actions - check if at least one action is present
292 * @exts: tc filter extensions handle
294 * Returns true if at least one action is present.
296 static inline bool tcf_exts_has_actions(struct tcf_exts *exts)
298 #ifdef CONFIG_NET_CLS_ACT
299 return exts->nr_actions;
306 * tcf_exts_exec - execute tc filter extensions
307 * @skb: socket buffer
308 * @exts: tc filter extensions handle
309 * @res: desired result
311 * Executes all configured extensions. Returns TC_ACT_OK on a normal execution,
312 * a negative number if the filter must be considered unmatched or
313 * a positive action code (TC_ACT_*) which must be returned to the
317 tcf_exts_exec(struct sk_buff *skb, struct tcf_exts *exts,
318 struct tcf_result *res)
320 #ifdef CONFIG_NET_CLS_ACT
321 return tcf_action_exec(skb, exts->actions, exts->nr_actions, res);
326 int tcf_exts_validate(struct net *net, struct tcf_proto *tp,
327 struct nlattr **tb, struct nlattr *rate_tlv,
328 struct tcf_exts *exts, bool ovr, bool rtnl_held,
329 struct netlink_ext_ack *extack);
330 void tcf_exts_destroy(struct tcf_exts *exts);
331 void tcf_exts_change(struct tcf_exts *dst, struct tcf_exts *src);
332 int tcf_exts_dump(struct sk_buff *skb, struct tcf_exts *exts);
333 int tcf_exts_dump_stats(struct sk_buff *skb, struct tcf_exts *exts);
336 * struct tcf_pkt_info - packet information
338 struct tcf_pkt_info {
343 #ifdef CONFIG_NET_EMATCH
345 struct tcf_ematch_ops;
348 * struct tcf_ematch - extended match (ematch)
350 * @matchid: identifier to allow userspace to reidentify a match
351 * @flags: flags specifying attributes and the relation to other matches
352 * @ops: the operations lookup table of the corresponding ematch module
353 * @datalen: length of the ematch specific configuration data
354 * @data: ematch specific data
357 struct tcf_ematch_ops * ops;
359 unsigned int datalen;
365 static inline int tcf_em_is_container(struct tcf_ematch *em)
370 static inline int tcf_em_is_simple(struct tcf_ematch *em)
372 return em->flags & TCF_EM_SIMPLE;
375 static inline int tcf_em_is_inverted(struct tcf_ematch *em)
377 return em->flags & TCF_EM_INVERT;
380 static inline int tcf_em_last_match(struct tcf_ematch *em)
382 return (em->flags & TCF_EM_REL_MASK) == TCF_EM_REL_END;
385 static inline int tcf_em_early_end(struct tcf_ematch *em, int result)
387 if (tcf_em_last_match(em))
390 if (result == 0 && em->flags & TCF_EM_REL_AND)
393 if (result != 0 && em->flags & TCF_EM_REL_OR)
400 * struct tcf_ematch_tree - ematch tree handle
402 * @hdr: ematch tree header supplied by userspace
403 * @matches: array of ematches
405 struct tcf_ematch_tree {
406 struct tcf_ematch_tree_hdr hdr;
407 struct tcf_ematch * matches;
412 * struct tcf_ematch_ops - ematch module operations
414 * @kind: identifier (kind) of this ematch module
415 * @datalen: length of expected configuration data (optional)
416 * @change: called during validation (optional)
417 * @match: called during ematch tree evaluation, must return 1/0
418 * @destroy: called during destroyage (optional)
419 * @dump: called during dumping process (optional)
420 * @owner: owner, must be set to THIS_MODULE
421 * @link: link to previous/next ematch module (internal use)
423 struct tcf_ematch_ops {
426 int (*change)(struct net *net, void *,
427 int, struct tcf_ematch *);
428 int (*match)(struct sk_buff *, struct tcf_ematch *,
429 struct tcf_pkt_info *);
430 void (*destroy)(struct tcf_ematch *);
431 int (*dump)(struct sk_buff *, struct tcf_ematch *);
432 struct module *owner;
433 struct list_head link;
436 int tcf_em_register(struct tcf_ematch_ops *);
437 void tcf_em_unregister(struct tcf_ematch_ops *);
438 int tcf_em_tree_validate(struct tcf_proto *, struct nlattr *,
439 struct tcf_ematch_tree *);
440 void tcf_em_tree_destroy(struct tcf_ematch_tree *);
441 int tcf_em_tree_dump(struct sk_buff *, struct tcf_ematch_tree *, int);
442 int __tcf_em_tree_match(struct sk_buff *, struct tcf_ematch_tree *,
443 struct tcf_pkt_info *);
446 * tcf_em_tree_match - evaulate an ematch tree
448 * @skb: socket buffer of the packet in question
449 * @tree: ematch tree to be used for evaluation
450 * @info: packet information examined by classifier
452 * This function matches @skb against the ematch tree in @tree by going
453 * through all ematches respecting their logic relations returning
454 * as soon as the result is obvious.
456 * Returns 1 if the ematch tree as-one matches, no ematches are configured
457 * or ematch is not enabled in the kernel, otherwise 0 is returned.
459 static inline int tcf_em_tree_match(struct sk_buff *skb,
460 struct tcf_ematch_tree *tree,
461 struct tcf_pkt_info *info)
463 if (tree->hdr.nmatches)
464 return __tcf_em_tree_match(skb, tree, info);
469 #define MODULE_ALIAS_TCF_EMATCH(kind) MODULE_ALIAS("ematch-kind-" __stringify(kind))
471 #else /* CONFIG_NET_EMATCH */
473 struct tcf_ematch_tree {
476 #define tcf_em_tree_validate(tp, tb, t) ((void)(t), 0)
477 #define tcf_em_tree_destroy(t) do { (void)(t); } while(0)
478 #define tcf_em_tree_dump(skb, t, tlv) (0)
479 #define tcf_em_tree_match(skb, t, info) ((void)(info), 1)
481 #endif /* CONFIG_NET_EMATCH */
483 static inline unsigned char * tcf_get_base_ptr(struct sk_buff *skb, int layer)
487 return skb_mac_header(skb);
488 case TCF_LAYER_NETWORK:
489 return skb_network_header(skb);
490 case TCF_LAYER_TRANSPORT:
491 return skb_transport_header(skb);
497 static inline int tcf_valid_offset(const struct sk_buff *skb,
498 const unsigned char *ptr, const int len)
500 return likely((ptr + len) <= skb_tail_pointer(skb) &&
502 (ptr <= (ptr + len)));
506 tcf_change_indev(struct net *net, struct nlattr *indev_tlv,
507 struct netlink_ext_ack *extack)
509 char indev[IFNAMSIZ];
510 struct net_device *dev;
512 if (nla_strlcpy(indev, indev_tlv, IFNAMSIZ) >= IFNAMSIZ) {
513 NL_SET_ERR_MSG(extack, "Interface name too long");
516 dev = __dev_get_by_name(net, indev);
523 tcf_match_indev(struct sk_buff *skb, int ifindex)
529 return ifindex == skb->skb_iif;
532 int tc_setup_flow_action(struct flow_action *flow_action,
533 const struct tcf_exts *exts);
534 int tc_setup_cb_call(struct tcf_block *block, enum tc_setup_type type,
535 void *type_data, bool err_stop);
536 unsigned int tcf_exts_num_actions(struct tcf_exts *exts);
538 struct tc_cls_u32_knode {
539 struct tcf_exts *exts;
540 struct tcf_result *res;
541 struct tc_u32_sel *sel;
549 struct tc_cls_u32_hnode {
552 unsigned int divisor;
555 enum tc_clsu32_command {
557 TC_CLSU32_REPLACE_KNODE,
558 TC_CLSU32_DELETE_KNODE,
560 TC_CLSU32_REPLACE_HNODE,
561 TC_CLSU32_DELETE_HNODE,
564 struct tc_cls_u32_offload {
565 struct flow_cls_common_offload common;
567 enum tc_clsu32_command command;
569 struct tc_cls_u32_knode knode;
570 struct tc_cls_u32_hnode hnode;
574 static inline bool tc_can_offload(const struct net_device *dev)
576 return dev->features & NETIF_F_HW_TC;
579 static inline bool tc_can_offload_extack(const struct net_device *dev,
580 struct netlink_ext_ack *extack)
582 bool can = tc_can_offload(dev);
585 NL_SET_ERR_MSG(extack, "TC offload is disabled on net device");
591 tc_cls_can_offload_and_chain0(const struct net_device *dev,
592 struct flow_cls_common_offload *common)
594 if (!tc_can_offload_extack(dev, common->extack))
596 if (common->chain_index) {
597 NL_SET_ERR_MSG(common->extack,
598 "Driver supports only offload of chain 0");
604 static inline bool tc_skip_hw(u32 flags)
606 return (flags & TCA_CLS_FLAGS_SKIP_HW) ? true : false;
609 static inline bool tc_skip_sw(u32 flags)
611 return (flags & TCA_CLS_FLAGS_SKIP_SW) ? true : false;
614 /* SKIP_HW and SKIP_SW are mutually exclusive flags. */
615 static inline bool tc_flags_valid(u32 flags)
617 if (flags & ~(TCA_CLS_FLAGS_SKIP_HW | TCA_CLS_FLAGS_SKIP_SW |
618 TCA_CLS_FLAGS_VERBOSE))
621 flags &= TCA_CLS_FLAGS_SKIP_HW | TCA_CLS_FLAGS_SKIP_SW;
622 if (!(flags ^ (TCA_CLS_FLAGS_SKIP_HW | TCA_CLS_FLAGS_SKIP_SW)))
628 static inline bool tc_in_hw(u32 flags)
630 return (flags & TCA_CLS_FLAGS_IN_HW) ? true : false;
634 tc_cls_common_offload_init(struct flow_cls_common_offload *cls_common,
635 const struct tcf_proto *tp, u32 flags,
636 struct netlink_ext_ack *extack)
638 cls_common->chain_index = tp->chain->index;
639 cls_common->protocol = tp->protocol;
640 cls_common->prio = tp->prio;
641 if (tc_skip_sw(flags) || flags & TCA_CLS_FLAGS_VERBOSE)
642 cls_common->extack = extack;
645 enum tc_matchall_command {
646 TC_CLSMATCHALL_REPLACE,
647 TC_CLSMATCHALL_DESTROY,
648 TC_CLSMATCHALL_STATS,
651 struct tc_cls_matchall_offload {
652 struct flow_cls_common_offload common;
653 enum tc_matchall_command command;
654 struct flow_rule *rule;
655 struct flow_stats stats;
656 unsigned long cookie;
659 enum tc_clsbpf_command {
664 struct tc_cls_bpf_offload {
665 struct flow_cls_common_offload common;
666 enum tc_clsbpf_command command;
667 struct tcf_exts *exts;
668 struct bpf_prog *prog;
669 struct bpf_prog *oldprog;
671 bool exts_integrated;
674 struct tc_mqprio_qopt_offload {
675 /* struct tc_mqprio_qopt must always be the first element */
676 struct tc_mqprio_qopt qopt;
680 u64 min_rate[TC_QOPT_MAX_QUEUE];
681 u64 max_rate[TC_QOPT_MAX_QUEUE];
684 /* This structure holds cookie structure that is passed from user
685 * to the kernel for actions and classifiers
693 struct tc_qopt_offload_stats {
694 struct gnet_stats_basic_packed *bstats;
695 struct gnet_stats_queue *qstats;
705 struct tc_mq_opt_offload_graft_params {
710 struct tc_mq_qopt_offload {
711 enum tc_mq_command command;
714 struct tc_qopt_offload_stats stats;
715 struct tc_mq_opt_offload_graft_params graft_params;
719 enum tc_red_command {
727 struct tc_red_qopt_offload_params {
734 struct gnet_stats_queue *qstats;
737 struct tc_red_qopt_offload {
738 enum tc_red_command command;
742 struct tc_red_qopt_offload_params set;
743 struct tc_qopt_offload_stats stats;
744 struct red_stats *xstats;
749 enum tc_gred_command {
755 struct tc_gred_vq_qopt_offload_params {
764 /* Only need backlog, see struct tc_prio_qopt_offload_params */
768 struct tc_gred_qopt_offload_params {
773 struct gnet_stats_queue *qstats;
774 struct tc_gred_vq_qopt_offload_params tab[MAX_DPs];
777 struct tc_gred_qopt_offload_stats {
778 struct gnet_stats_basic_packed bstats[MAX_DPs];
779 struct gnet_stats_queue qstats[MAX_DPs];
780 struct red_stats *xstats[MAX_DPs];
783 struct tc_gred_qopt_offload {
784 enum tc_gred_command command;
788 struct tc_gred_qopt_offload_params set;
789 struct tc_gred_qopt_offload_stats stats;
793 enum tc_prio_command {
800 struct tc_prio_qopt_offload_params {
802 u8 priomap[TC_PRIO_MAX + 1];
803 /* In case that a prio qdisc is offloaded and now is changed to a
804 * non-offloadedable config, it needs to update the backlog & qlen
805 * values to negate the HW backlog & qlen values (and only them).
807 struct gnet_stats_queue *qstats;
810 struct tc_prio_qopt_offload_graft_params {
815 struct tc_prio_qopt_offload {
816 enum tc_prio_command command;
820 struct tc_prio_qopt_offload_params replace_params;
821 struct tc_qopt_offload_stats stats;
822 struct tc_prio_qopt_offload_graft_params graft_params;
826 enum tc_root_command {
830 struct tc_root_qopt_offload {
831 enum tc_root_command command;