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 work_struct *work);
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 struct Qdisc *tcf_block_q(struct tcf_block *block)
119 static inline struct net_device *tcf_block_dev(struct tcf_block *block)
125 int tc_setup_cb_block_register(struct tcf_block *block, tc_setup_cb_t *cb,
132 void tc_setup_cb_block_unregister(struct tcf_block *block, tc_setup_cb_t *cb,
138 void *tcf_block_cb_priv(struct tcf_block_cb *block_cb)
144 struct tcf_block_cb *tcf_block_cb_lookup(struct tcf_block *block,
145 tc_setup_cb_t *cb, void *cb_ident)
151 void tcf_block_cb_incref(struct tcf_block_cb *block_cb)
156 unsigned int tcf_block_cb_decref(struct tcf_block_cb *block_cb)
162 struct tcf_block_cb *__tcf_block_cb_register(struct tcf_block *block,
163 tc_setup_cb_t *cb, void *cb_ident,
170 int tcf_block_cb_register(struct tcf_block *block,
171 tc_setup_cb_t *cb, void *cb_ident,
178 void __tcf_block_cb_unregister(struct tcf_block_cb *block_cb)
183 void tcf_block_cb_unregister(struct tcf_block *block,
184 tc_setup_cb_t *cb, void *cb_ident)
188 static inline int tcf_classify(struct sk_buff *skb, const struct tcf_proto *tp,
189 struct tcf_result *res, bool compat_mode)
191 return TC_ACT_UNSPEC;
195 static inline unsigned long
196 __cls_set_class(unsigned long *clp, unsigned long cl)
198 return xchg(clp, cl);
201 static inline unsigned long
202 cls_set_class(struct Qdisc *q, unsigned long *clp, unsigned long cl)
204 unsigned long old_cl;
207 old_cl = __cls_set_class(clp, cl);
213 tcf_bind_filter(struct tcf_proto *tp, struct tcf_result *r, unsigned long base)
215 struct Qdisc *q = tp->chain->block->q;
218 /* Check q as it is not set for shared blocks. In that case,
219 * setting class is not supported.
223 cl = q->ops->cl_ops->bind_tcf(q, base, r->classid);
224 cl = cls_set_class(q, &r->class, cl);
226 q->ops->cl_ops->unbind_tcf(q, cl);
230 tcf_unbind_filter(struct tcf_proto *tp, struct tcf_result *r)
232 struct Qdisc *q = tp->chain->block->q;
237 if ((cl = __cls_set_class(&r->class, 0)) != 0)
238 q->ops->cl_ops->unbind_tcf(q, cl);
242 #ifdef CONFIG_NET_CLS_ACT
243 __u32 type; /* for backward compat(TCA_OLD_COMPAT) */
245 struct tc_action **actions;
248 /* Map to export classifier specific extension TLV types to the
249 * generic extensions API. Unsupported extensions must be set to 0.
255 static inline int tcf_exts_init(struct tcf_exts *exts, int action, int police)
257 #ifdef CONFIG_NET_CLS_ACT
259 exts->nr_actions = 0;
261 exts->actions = kcalloc(TCA_ACT_MAX_PRIO, sizeof(struct tc_action *),
266 exts->action = action;
267 exts->police = police;
271 /* Return false if the netns is being destroyed in cleanup_net(). Callers
272 * need to do cleanup synchronously in this case, otherwise may race with
273 * tc_action_net_exit(). Return true for other cases.
275 static inline bool tcf_exts_get_net(struct tcf_exts *exts)
277 #ifdef CONFIG_NET_CLS_ACT
278 exts->net = maybe_get_net(exts->net);
279 return exts->net != NULL;
285 static inline void tcf_exts_put_net(struct tcf_exts *exts)
287 #ifdef CONFIG_NET_CLS_ACT
293 static inline void tcf_exts_to_list(const struct tcf_exts *exts,
294 struct list_head *actions)
296 #ifdef CONFIG_NET_CLS_ACT
299 for (i = 0; i < exts->nr_actions; i++) {
300 struct tc_action *a = exts->actions[i];
302 list_add_tail(&a->list, actions);
308 tcf_exts_stats_update(const struct tcf_exts *exts,
309 u64 bytes, u64 packets, u64 lastuse)
311 #ifdef CONFIG_NET_CLS_ACT
316 for (i = 0; i < exts->nr_actions; i++) {
317 struct tc_action *a = exts->actions[i];
319 tcf_action_stats_update(a, bytes, packets, lastuse);
327 * tcf_exts_has_actions - check if at least one action is present
328 * @exts: tc filter extensions handle
330 * Returns true if at least one action is present.
332 static inline bool tcf_exts_has_actions(struct tcf_exts *exts)
334 #ifdef CONFIG_NET_CLS_ACT
335 return exts->nr_actions;
342 * tcf_exts_has_one_action - check if exactly one action is present
343 * @exts: tc filter extensions handle
345 * Returns true if exactly one action is present.
347 static inline bool tcf_exts_has_one_action(struct tcf_exts *exts)
349 #ifdef CONFIG_NET_CLS_ACT
350 return exts->nr_actions == 1;
357 * tcf_exts_exec - execute tc filter extensions
358 * @skb: socket buffer
359 * @exts: tc filter extensions handle
360 * @res: desired result
362 * Executes all configured extensions. Returns TC_ACT_OK on a normal execution,
363 * a negative number if the filter must be considered unmatched or
364 * a positive action code (TC_ACT_*) which must be returned to the
368 tcf_exts_exec(struct sk_buff *skb, struct tcf_exts *exts,
369 struct tcf_result *res)
371 #ifdef CONFIG_NET_CLS_ACT
372 return tcf_action_exec(skb, exts->actions, exts->nr_actions, res);
377 int tcf_exts_validate(struct net *net, struct tcf_proto *tp,
378 struct nlattr **tb, struct nlattr *rate_tlv,
379 struct tcf_exts *exts, bool ovr,
380 struct netlink_ext_ack *extack);
381 void tcf_exts_destroy(struct tcf_exts *exts);
382 void tcf_exts_change(struct tcf_exts *dst, struct tcf_exts *src);
383 int tcf_exts_dump(struct sk_buff *skb, struct tcf_exts *exts);
384 int tcf_exts_dump_stats(struct sk_buff *skb, struct tcf_exts *exts);
387 * struct tcf_pkt_info - packet information
389 struct tcf_pkt_info {
394 #ifdef CONFIG_NET_EMATCH
396 struct tcf_ematch_ops;
399 * struct tcf_ematch - extended match (ematch)
401 * @matchid: identifier to allow userspace to reidentify a match
402 * @flags: flags specifying attributes and the relation to other matches
403 * @ops: the operations lookup table of the corresponding ematch module
404 * @datalen: length of the ematch specific configuration data
405 * @data: ematch specific data
408 struct tcf_ematch_ops * ops;
410 unsigned int datalen;
416 static inline int tcf_em_is_container(struct tcf_ematch *em)
421 static inline int tcf_em_is_simple(struct tcf_ematch *em)
423 return em->flags & TCF_EM_SIMPLE;
426 static inline int tcf_em_is_inverted(struct tcf_ematch *em)
428 return em->flags & TCF_EM_INVERT;
431 static inline int tcf_em_last_match(struct tcf_ematch *em)
433 return (em->flags & TCF_EM_REL_MASK) == TCF_EM_REL_END;
436 static inline int tcf_em_early_end(struct tcf_ematch *em, int result)
438 if (tcf_em_last_match(em))
441 if (result == 0 && em->flags & TCF_EM_REL_AND)
444 if (result != 0 && em->flags & TCF_EM_REL_OR)
451 * struct tcf_ematch_tree - ematch tree handle
453 * @hdr: ematch tree header supplied by userspace
454 * @matches: array of ematches
456 struct tcf_ematch_tree {
457 struct tcf_ematch_tree_hdr hdr;
458 struct tcf_ematch * matches;
463 * struct tcf_ematch_ops - ematch module operations
465 * @kind: identifier (kind) of this ematch module
466 * @datalen: length of expected configuration data (optional)
467 * @change: called during validation (optional)
468 * @match: called during ematch tree evaluation, must return 1/0
469 * @destroy: called during destroyage (optional)
470 * @dump: called during dumping process (optional)
471 * @owner: owner, must be set to THIS_MODULE
472 * @link: link to previous/next ematch module (internal use)
474 struct tcf_ematch_ops {
477 int (*change)(struct net *net, void *,
478 int, struct tcf_ematch *);
479 int (*match)(struct sk_buff *, struct tcf_ematch *,
480 struct tcf_pkt_info *);
481 void (*destroy)(struct tcf_ematch *);
482 int (*dump)(struct sk_buff *, struct tcf_ematch *);
483 struct module *owner;
484 struct list_head link;
487 int tcf_em_register(struct tcf_ematch_ops *);
488 void tcf_em_unregister(struct tcf_ematch_ops *);
489 int tcf_em_tree_validate(struct tcf_proto *, struct nlattr *,
490 struct tcf_ematch_tree *);
491 void tcf_em_tree_destroy(struct tcf_ematch_tree *);
492 int tcf_em_tree_dump(struct sk_buff *, struct tcf_ematch_tree *, int);
493 int __tcf_em_tree_match(struct sk_buff *, struct tcf_ematch_tree *,
494 struct tcf_pkt_info *);
497 * tcf_em_tree_match - evaulate an ematch tree
499 * @skb: socket buffer of the packet in question
500 * @tree: ematch tree to be used for evaluation
501 * @info: packet information examined by classifier
503 * This function matches @skb against the ematch tree in @tree by going
504 * through all ematches respecting their logic relations returning
505 * as soon as the result is obvious.
507 * Returns 1 if the ematch tree as-one matches, no ematches are configured
508 * or ematch is not enabled in the kernel, otherwise 0 is returned.
510 static inline int tcf_em_tree_match(struct sk_buff *skb,
511 struct tcf_ematch_tree *tree,
512 struct tcf_pkt_info *info)
514 if (tree->hdr.nmatches)
515 return __tcf_em_tree_match(skb, tree, info);
520 #define MODULE_ALIAS_TCF_EMATCH(kind) MODULE_ALIAS("ematch-kind-" __stringify(kind))
522 #else /* CONFIG_NET_EMATCH */
524 struct tcf_ematch_tree {
527 #define tcf_em_tree_validate(tp, tb, t) ((void)(t), 0)
528 #define tcf_em_tree_destroy(t) do { (void)(t); } while(0)
529 #define tcf_em_tree_dump(skb, t, tlv) (0)
530 #define tcf_em_tree_match(skb, t, info) ((void)(info), 1)
532 #endif /* CONFIG_NET_EMATCH */
534 static inline unsigned char * tcf_get_base_ptr(struct sk_buff *skb, int layer)
538 return skb_mac_header(skb);
539 case TCF_LAYER_NETWORK:
540 return skb_network_header(skb);
541 case TCF_LAYER_TRANSPORT:
542 return skb_transport_header(skb);
548 static inline int tcf_valid_offset(const struct sk_buff *skb,
549 const unsigned char *ptr, const int len)
551 return likely((ptr + len) <= skb_tail_pointer(skb) &&
553 (ptr <= (ptr + len)));
556 #ifdef CONFIG_NET_CLS_IND
557 #include <net/net_namespace.h>
560 tcf_change_indev(struct net *net, struct nlattr *indev_tlv,
561 struct netlink_ext_ack *extack)
563 char indev[IFNAMSIZ];
564 struct net_device *dev;
566 if (nla_strlcpy(indev, indev_tlv, IFNAMSIZ) >= IFNAMSIZ) {
567 NL_SET_ERR_MSG(extack, "Interface name too long");
570 dev = __dev_get_by_name(net, indev);
577 tcf_match_indev(struct sk_buff *skb, int ifindex)
583 return ifindex == skb->skb_iif;
585 #endif /* CONFIG_NET_CLS_IND */
587 int tc_setup_cb_call(struct tcf_block *block, struct tcf_exts *exts,
588 enum tc_setup_type type, void *type_data, bool err_stop);
590 enum tc_block_command {
595 struct tc_block_offload {
596 enum tc_block_command command;
597 enum tcf_block_binder_type binder_type;
598 struct tcf_block *block;
601 struct tc_cls_common_offload {
605 struct netlink_ext_ack *extack;
608 struct tc_cls_u32_knode {
609 struct tcf_exts *exts;
610 struct tc_u32_sel *sel;
618 struct tc_cls_u32_hnode {
621 unsigned int divisor;
624 enum tc_clsu32_command {
626 TC_CLSU32_REPLACE_KNODE,
627 TC_CLSU32_DELETE_KNODE,
629 TC_CLSU32_REPLACE_HNODE,
630 TC_CLSU32_DELETE_HNODE,
633 struct tc_cls_u32_offload {
634 struct tc_cls_common_offload common;
636 enum tc_clsu32_command command;
638 struct tc_cls_u32_knode knode;
639 struct tc_cls_u32_hnode hnode;
643 static inline bool tc_can_offload(const struct net_device *dev)
645 return dev->features & NETIF_F_HW_TC;
648 static inline bool tc_can_offload_extack(const struct net_device *dev,
649 struct netlink_ext_ack *extack)
651 bool can = tc_can_offload(dev);
654 NL_SET_ERR_MSG(extack, "TC offload is disabled on net device");
660 tc_cls_can_offload_and_chain0(const struct net_device *dev,
661 struct tc_cls_common_offload *common)
663 if (!tc_can_offload_extack(dev, common->extack))
665 if (common->chain_index) {
666 NL_SET_ERR_MSG(common->extack,
667 "Driver supports only offload of chain 0");
673 static inline bool tc_skip_hw(u32 flags)
675 return (flags & TCA_CLS_FLAGS_SKIP_HW) ? true : false;
678 static inline bool tc_skip_sw(u32 flags)
680 return (flags & TCA_CLS_FLAGS_SKIP_SW) ? true : false;
683 /* SKIP_HW and SKIP_SW are mutually exclusive flags. */
684 static inline bool tc_flags_valid(u32 flags)
686 if (flags & ~(TCA_CLS_FLAGS_SKIP_HW | TCA_CLS_FLAGS_SKIP_SW))
689 if (!(flags ^ (TCA_CLS_FLAGS_SKIP_HW | TCA_CLS_FLAGS_SKIP_SW)))
695 static inline bool tc_in_hw(u32 flags)
697 return (flags & TCA_CLS_FLAGS_IN_HW) ? true : false;
701 tc_cls_common_offload_init(struct tc_cls_common_offload *cls_common,
702 const struct tcf_proto *tp, u32 flags,
703 struct netlink_ext_ack *extack)
705 cls_common->chain_index = tp->chain->index;
706 cls_common->protocol = tp->protocol;
707 cls_common->prio = tp->prio;
708 if (tc_skip_sw(flags))
709 cls_common->extack = extack;
713 TC_CLSFLOWER_REPLACE,
714 TC_CLSFLOWER_DESTROY,
718 struct tc_cls_flower_offload {
719 struct tc_cls_common_offload common;
720 enum tc_fl_command command;
721 unsigned long cookie;
722 struct flow_dissector *dissector;
723 struct fl_flow_key *mask;
724 struct fl_flow_key *key;
725 struct tcf_exts *exts;
729 enum tc_matchall_command {
730 TC_CLSMATCHALL_REPLACE,
731 TC_CLSMATCHALL_DESTROY,
734 struct tc_cls_matchall_offload {
735 struct tc_cls_common_offload common;
736 enum tc_matchall_command command;
737 struct tcf_exts *exts;
738 unsigned long cookie;
741 enum tc_clsbpf_command {
746 struct tc_cls_bpf_offload {
747 struct tc_cls_common_offload common;
748 enum tc_clsbpf_command command;
749 struct tcf_exts *exts;
750 struct bpf_prog *prog;
751 struct bpf_prog *oldprog;
753 bool exts_integrated;
756 struct tc_mqprio_qopt_offload {
757 /* struct tc_mqprio_qopt must always be the first element */
758 struct tc_mqprio_qopt qopt;
762 u64 min_rate[TC_QOPT_MAX_QUEUE];
763 u64 max_rate[TC_QOPT_MAX_QUEUE];
766 /* This structure holds cookie structure that is passed from user
767 * to the kernel for actions and classifiers
774 struct tc_qopt_offload_stats {
775 struct gnet_stats_basic_packed *bstats;
776 struct gnet_stats_queue *qstats;
779 enum tc_red_command {
786 struct tc_red_qopt_offload_params {
791 struct gnet_stats_queue *qstats;
794 struct tc_red_qopt_offload {
795 enum tc_red_command command;
799 struct tc_red_qopt_offload_params set;
800 struct tc_qopt_offload_stats stats;
801 struct red_stats *xstats;
805 enum tc_prio_command {
812 struct tc_prio_qopt_offload_params {
814 u8 priomap[TC_PRIO_MAX + 1];
815 /* In case that a prio qdisc is offloaded and now is changed to a
816 * non-offloadedable config, it needs to update the backlog & qlen
817 * values to negate the HW backlog & qlen values (and only them).
819 struct gnet_stats_queue *qstats;
822 struct tc_prio_qopt_offload_graft_params {
827 struct tc_prio_qopt_offload {
828 enum tc_prio_command command;
832 struct tc_prio_qopt_offload_params replace_params;
833 struct tc_qopt_offload_stats stats;
834 struct tc_prio_qopt_offload_graft_params graft_params;