forked from ~ljy/RK356X_SDK_RELEASE
blame | history | patch | commit | commitdiff
hc
2024-05-10 23fa18eaa71266feff7ba8d83022d9e1cc83c65a 
 kernel/net/sched/cls_u32.c
....@@ -1,10 +1,6 @@
1
+// SPDX-License-Identifier: GPL-2.0-or-later
12 /*
23  * net/sched/cls_u32.c	Ugly (or Universal) 32bit key Packet Classifier.
3
- *
4
- *		This program is free software; you can redistribute it and/or
5
- *		modify it under the terms of the GNU General Public License
6
- *		as published by the Free Software Foundation; either version
7
- *		2 of the License, or (at your option) any later version.
84  *
95  * Authors:	Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
106  *
....@@ -23,9 +19,6 @@
2319  *	It is especially useful for link sharing combined with QoS;
2420  *	pure RSVP doesn't need such a general approach and can use
2521  *	much simpler (and faster) schemes, sort of cls_rsvp.c.
26
- *
27
- *	JHS: We should remove the CONFIG_NET_CLS_IND from here
28
- *	eventually when the meta match extension is made available
2922  *
3023  *	nfmark match added by Catalin(ux aka Dino) BOIE <catab at umbrella.ro>
3124  */
....@@ -52,9 +45,7 @@
5245 	u32			handle;
5346 	struct tc_u_hnode __rcu	*ht_up;
5447 	struct tcf_exts		exts;
55
-#ifdef CONFIG_NET_CLS_IND
5648 	int			ifindex;
57
-#endif
5849 	u8			fshift;
5950 	struct tcf_result	res;
6051 	struct tc_u_hnode __rcu	*ht_down;
....@@ -68,7 +59,6 @@
6859 	u32			mask;
6960 	u32 __percpu		*pcpu_success;
7061 #endif
71
-	struct tcf_proto	*tp;
7262 	struct rcu_work		rwork;
7363 	/* The 'sel' field MUST be the last field in structure to allow for
7464 	 * tc_u32_keys allocated at end of structure.
....@@ -80,16 +70,16 @@
8070 	struct tc_u_hnode __rcu	*next;
8171 	u32			handle;
8272 	u32			prio;
83
-	struct tc_u_common	*tp_c;
8473 	int			refcnt;
8574 	unsigned int		divisor;
8675 	struct idr		handle_idr;
76
+	bool			is_root;
8777 	struct rcu_head		rcu;
8878 	u32			flags;
8979 	/* The 'ht' field MUST be the last field in structure to allow for
9080 	 * more entries allocated at end of structure.
9181 	 */
92
-	struct tc_u_knode __rcu	*ht[1];
82
+	struct tc_u_knode __rcu	*ht[];
9383 };
9484 
9585 struct tc_u_common {
....@@ -98,7 +88,7 @@
9888 	int			refcnt;
9989 	struct idr		handle_idr;
10090 	struct hlist_node	hnode;
101
-	struct rcu_head		rcu;
91
+	long			knodes;
10292 };
10393 
10494 static inline unsigned int u32_hash_fold(__be32 key,
....@@ -181,12 +171,10 @@
181171 			if (n->sel.flags & TC_U32_TERMINAL) {
182172 
183173 				*res = n->res;
184
-#ifdef CONFIG_NET_CLS_IND
185174 				if (!tcf_match_indev(skb, n->ifindex)) {
186175 					n = rcu_dereference_bh(n->next);
187176 					goto next_knode;
188177 				}
189
-#endif
190178 #ifdef CONFIG_CLS_U32_PERF
191179 				__this_cpu_inc(n->pf->rhit);
192180 #endif
....@@ -344,19 +332,16 @@
344332 		return block->q;
345333 }
346334 
347
-static unsigned int tc_u_hash(const struct tcf_proto *tp)
335
+static struct hlist_head *tc_u_hash(void *key)
348336 {
349
-	return hash_ptr(tc_u_common_ptr(tp), U32_HASH_SHIFT);
337
+	return tc_u_common_hash + hash_ptr(key, U32_HASH_SHIFT);
350338 }
351339 
352
-static struct tc_u_common *tc_u_common_find(const struct tcf_proto *tp)
340
+static struct tc_u_common *tc_u_common_find(void *key)
353341 {
354342 	struct tc_u_common *tc;
355
-	unsigned int h;
356
-
357
-	h = tc_u_hash(tp);
358
-	hlist_for_each_entry(tc, &tc_u_common_hash[h], hnode) {
359
-		if (tc->ptr == tc_u_common_ptr(tp))
343
+	hlist_for_each_entry(tc, tc_u_hash(key), hnode) {
344
+		if (tc->ptr == key)
360345 			return tc;
361346 	}
362347 	return NULL;
....@@ -365,38 +350,35 @@
365350 static int u32_init(struct tcf_proto *tp)
366351 {
367352 	struct tc_u_hnode *root_ht;
368
-	struct tc_u_common *tp_c;
369
-	unsigned int h;
353
+	void *key = tc_u_common_ptr(tp);
354
+	struct tc_u_common *tp_c = tc_u_common_find(key);
370355 
371
-	tp_c = tc_u_common_find(tp);
372
-
373
-	root_ht = kzalloc(sizeof(*root_ht), GFP_KERNEL);
356
+	root_ht = kzalloc(struct_size(root_ht, ht, 1), GFP_KERNEL);
374357 	if (root_ht == NULL)
375358 		return -ENOBUFS;
376359 
377360 	root_ht->refcnt++;
378361 	root_ht->handle = tp_c ? gen_new_htid(tp_c, root_ht) : 0x80000000;
379362 	root_ht->prio = tp->prio;
363
+	root_ht->is_root = true;
380364 	idr_init(&root_ht->handle_idr);
381365 
382366 	if (tp_c == NULL) {
383
-		tp_c = kzalloc(sizeof(*tp_c), GFP_KERNEL);
367
+		tp_c = kzalloc(struct_size(tp_c, hlist->ht, 1), GFP_KERNEL);
384368 		if (tp_c == NULL) {
385369 			kfree(root_ht);
386370 			return -ENOBUFS;
387371 		}
388
-		tp_c->ptr = tc_u_common_ptr(tp);
372
+		tp_c->ptr = key;
389373 		INIT_HLIST_NODE(&tp_c->hnode);
390374 		idr_init(&tp_c->handle_idr);
391375 
392
-		h = tc_u_hash(tp);
393
-		hlist_add_head(&tp_c->hnode, &tc_u_common_hash[h]);
376
+		hlist_add_head(&tp_c->hnode, tc_u_hash(key));
394377 	}
395378 
396379 	tp_c->refcnt++;
397380 	RCU_INIT_POINTER(root_ht->next, tp_c->hlist);
398381 	rcu_assign_pointer(tp_c->hlist, root_ht);
399
-	root_ht->tp_c = tp_c;
400382 
401383 	root_ht->refcnt++;
402384 	rcu_assign_pointer(tp->root, root_ht);
....@@ -404,15 +386,19 @@
404386 	return 0;
405387 }
406388 
407
-static int u32_destroy_key(struct tcf_proto *tp, struct tc_u_knode *n,
408
-			   bool free_pf)
389
+static void __u32_destroy_key(struct tc_u_knode *n)
409390 {
410391 	struct tc_u_hnode *ht = rtnl_dereference(n->ht_down);
411392 
412393 	tcf_exts_destroy(&n->exts);
413
-	tcf_exts_put_net(&n->exts);
414394 	if (ht && --ht->refcnt == 0)
415395 		kfree(ht);
396
+	kfree(n);
397
+}
398
+
399
+static void u32_destroy_key(struct tc_u_knode *n, bool free_pf)
400
+{
401
+	tcf_exts_put_net(&n->exts);
416402 #ifdef CONFIG_CLS_U32_PERF
417403 	if (free_pf)
418404 		free_percpu(n->pf);
....@@ -421,8 +407,7 @@
421407 	if (free_pf)
422408 		free_percpu(n->pcpu_success);
423409 #endif
424
-	kfree(n);
425
-	return 0;
410
+	__u32_destroy_key(n);
426411 }
427412 
428413 /* u32_delete_key_rcu should be called when free'ing a copied
....@@ -439,7 +424,7 @@
439424 					      struct tc_u_knode,
440425 					      rwork);
441426 	rtnl_lock();
442
-	u32_destroy_key(key->tp, key, false);
427
+	u32_destroy_key(key, false);
443428 	rtnl_unlock();
444429 }
445430 
....@@ -456,12 +441,13 @@
456441 					      struct tc_u_knode,
457442 					      rwork);
458443 	rtnl_lock();
459
-	u32_destroy_key(key->tp, key, true);
444
+	u32_destroy_key(key, true);
460445 	rtnl_unlock();
461446 }
462447 
463448 static int u32_delete_key(struct tcf_proto *tp, struct tc_u_knode *key)
464449 {
450
+	struct tc_u_common *tp_c = tp->data;
465451 	struct tc_u_knode __rcu **kp;
466452 	struct tc_u_knode *pkp;
467453 	struct tc_u_hnode *ht = rtnl_dereference(key->ht_up);
....@@ -472,6 +458,7 @@
472458 		     kp = &pkp->next, pkp = rtnl_dereference(*kp)) {
473459 			if (pkp == key) {
474460 				RCU_INIT_POINTER(*kp, key->next);
461
+				tp_c->knodes--;
475462 
476463 				tcf_unbind_filter(tp, &key->res);
477464 				idr_remove(&ht->handle_idr, key->handle);
....@@ -497,7 +484,7 @@
497484 	cls_u32.hnode.handle = h->handle;
498485 	cls_u32.hnode.prio = h->prio;
499486 
500
-	tc_setup_cb_call(block, NULL, TC_SETUP_CLSU32, &cls_u32, false);
487
+	tc_setup_cb_call(block, TC_SETUP_CLSU32, &cls_u32, false, true);
501488 }
502489 
503490 static int u32_replace_hw_hnode(struct tcf_proto *tp, struct tc_u_hnode *h,
....@@ -515,7 +502,7 @@
515502 	cls_u32.hnode.handle = h->handle;
516503 	cls_u32.hnode.prio = h->prio;
517504 
518
-	err = tc_setup_cb_call(block, NULL, TC_SETUP_CLSU32, &cls_u32, skip_sw);
505
+	err = tc_setup_cb_call(block, TC_SETUP_CLSU32, &cls_u32, skip_sw, true);
519506 	if (err < 0) {
520507 		u32_clear_hw_hnode(tp, h, NULL);
521508 		return err;
....@@ -539,8 +526,8 @@
539526 	cls_u32.command = TC_CLSU32_DELETE_KNODE;
540527 	cls_u32.knode.handle = n->handle;
541528 
542
-	tc_setup_cb_call(block, NULL, TC_SETUP_CLSU32, &cls_u32, false);
543
-	tcf_block_offload_dec(block, &n->flags);
529
+	tc_setup_cb_destroy(block, tp, TC_SETUP_CLSU32, &cls_u32, false,
530
+			    &n->flags, &n->in_hw_count, true);
544531 }
545532 
546533 static int u32_replace_hw_knode(struct tcf_proto *tp, struct tc_u_knode *n,
....@@ -564,17 +551,16 @@
564551 	cls_u32.knode.mask = 0;
565552 #endif
566553 	cls_u32.knode.sel = &n->sel;
554
+	cls_u32.knode.res = &n->res;
567555 	cls_u32.knode.exts = &n->exts;
568556 	if (n->ht_down)
569557 		cls_u32.knode.link_handle = ht->handle;
570558 
571
-	err = tc_setup_cb_call(block, NULL, TC_SETUP_CLSU32, &cls_u32, skip_sw);
572
-	if (err < 0) {
559
+	err = tc_setup_cb_add(block, tp, TC_SETUP_CLSU32, &cls_u32, skip_sw,
560
+			      &n->flags, &n->in_hw_count, true);
561
+	if (err) {
573562 		u32_remove_hw_knode(tp, n, NULL);
574563 		return err;
575
-	} else if (err > 0) {
576
-		n->in_hw_count = err;
577
-		tcf_block_offload_inc(block, &n->flags);
578564 	}
579565 
580566 	if (skip_sw && !(n->flags & TCA_CLS_FLAGS_IN_HW))
....@@ -586,6 +572,7 @@
586572 static void u32_clear_hnode(struct tcf_proto *tp, struct tc_u_hnode *ht,
587573 			    struct netlink_ext_ack *extack)
588574 {
575
+	struct tc_u_common *tp_c = tp->data;
589576 	struct tc_u_knode *n;
590577 	unsigned int h;
591578 
....@@ -593,13 +580,14 @@
593580 		while ((n = rtnl_dereference(ht->ht[h])) != NULL) {
594581 			RCU_INIT_POINTER(ht->ht[h],
595582 					 rtnl_dereference(n->next));
583
+			tp_c->knodes--;
596584 			tcf_unbind_filter(tp, &n->res);
597585 			u32_remove_hw_knode(tp, n, extack);
598586 			idr_remove(&ht->handle_idr, n->handle);
599587 			if (tcf_exts_get_net(&n->exts))
600588 				tcf_queue_work(&n->rwork, u32_delete_key_freepf_work);
601589 			else
602
-				u32_destroy_key(n->tp, n, true);
590
+				u32_destroy_key(n, true);
603591 		}
604592 	}
605593 }
....@@ -632,18 +620,8 @@
632620 	return -ENOENT;
633621 }
634622 
635
-static bool ht_empty(struct tc_u_hnode *ht)
636
-{
637
-	unsigned int h;
638
-
639
-	for (h = 0; h <= ht->divisor; h++)
640
-		if (rcu_access_pointer(ht->ht[h]))
641
-			return false;
642
-
643
-	return true;
644
-}
645
-
646
-static void u32_destroy(struct tcf_proto *tp, struct netlink_ext_ack *extack)
623
+static void u32_destroy(struct tcf_proto *tp, bool rtnl_held,
624
+			struct netlink_ext_ack *extack)
647625 {
648626 	struct tc_u_common *tp_c = tp->data;
649627 	struct tc_u_hnode *root_ht = rtnl_dereference(tp->root);
....@@ -677,15 +655,11 @@
677655 }
678656 
679657 static int u32_delete(struct tcf_proto *tp, void *arg, bool *last,
680
-		      struct netlink_ext_ack *extack)
658
+		      bool rtnl_held, struct netlink_ext_ack *extack)
681659 {
682660 	struct tc_u_hnode *ht = arg;
683
-	struct tc_u_hnode *root_ht = rtnl_dereference(tp->root);
684661 	struct tc_u_common *tp_c = tp->data;
685662 	int ret = 0;
686
-
687
-	if (ht == NULL)
688
-		goto out;
689663 
690664 	if (TC_U32_KEY(ht->handle)) {
691665 		u32_remove_hw_knode(tp, (struct tc_u_knode *)ht, extack);
....@@ -693,7 +667,7 @@
693667 		goto out;
694668 	}
695669 
696
-	if (root_ht == ht) {
670
+	if (ht->is_root) {
697671 		NL_SET_ERR_MSG_MOD(extack, "Not allowed to delete root node");
698672 		return -EINVAL;
699673 	}
....@@ -706,38 +680,7 @@
706680 	}
707681 
708682 out:
709
-	*last = true;
710
-	if (root_ht) {
711
-		if (root_ht->refcnt > 2) {
712
-			*last = false;
713
-			goto ret;
714
-		}
715
-		if (root_ht->refcnt == 2) {
716
-			if (!ht_empty(root_ht)) {
717
-				*last = false;
718
-				goto ret;
719
-			}
720
-		}
721
-	}
722
-
723
-	if (tp_c->refcnt > 1) {
724
-		*last = false;
725
-		goto ret;
726
-	}
727
-
728
-	if (tp_c->refcnt == 1) {
729
-		struct tc_u_hnode *ht;
730
-
731
-		for (ht = rtnl_dereference(tp_c->hlist);
732
-		     ht;
733
-		     ht = rtnl_dereference(ht->next))
734
-			if (!ht_empty(ht)) {
735
-				*last = false;
736
-				break;
737
-			}
738
-	}
739
-
740
-ret:
683
+	*last = tp_c->refcnt == 1 && tp_c->knodes == 0;
741684 	return ret;
742685 }
743686 
....@@ -768,16 +711,22 @@
768711 };
769712 
770713 static int u32_set_parms(struct net *net, struct tcf_proto *tp,
771
-			 unsigned long base, struct tc_u_hnode *ht,
714
+			 unsigned long base,
772715 			 struct tc_u_knode *n, struct nlattr **tb,
773716 			 struct nlattr *est, bool ovr,
774717 			 struct netlink_ext_ack *extack)
775718 {
776
-	int err;
719
+	int err, ifindex = -1;
777720 
778
-	err = tcf_exts_validate(net, tp, tb, est, &n->exts, ovr, extack);
721
+	err = tcf_exts_validate(net, tp, tb, est, &n->exts, ovr, true, extack);
779722 	if (err < 0)
780723 		return err;
724
+
725
+	if (tb[TCA_U32_INDEV]) {
726
+		ifindex = tcf_change_indev(net, tb[TCA_U32_INDEV], extack);
727
+		if (ifindex < 0)
728
+			return -EINVAL;
729
+	}
781730 
782731 	if (tb[TCA_U32_LINK]) {
783732 		u32 handle = nla_get_u32(tb[TCA_U32_LINK]);
....@@ -789,10 +738,14 @@
789738 		}
790739 
791740 		if (handle) {
792
-			ht_down = u32_lookup_ht(ht->tp_c, handle);
741
+			ht_down = u32_lookup_ht(tp->data, handle);
793742 
794743 			if (!ht_down) {
795744 				NL_SET_ERR_MSG_MOD(extack, "Link hash table not found");
745
+				return -EINVAL;
746
+			}
747
+			if (ht_down->is_root) {
748
+				NL_SET_ERR_MSG_MOD(extack, "Not linking to root node");
796749 				return -EINVAL;
797750 			}
798751 			ht_down->refcnt++;
....@@ -809,15 +762,9 @@
809762 		tcf_bind_filter(tp, &n->res, base);
810763 	}
811764 
812
-#ifdef CONFIG_NET_CLS_IND
813
-	if (tb[TCA_U32_INDEV]) {
814
-		int ret;
815
-		ret = tcf_change_indev(net, tb[TCA_U32_INDEV], extack);
816
-		if (ret < 0)
817
-			return -EINVAL;
818
-		n->ifindex = ret;
819
-	}
820
-#endif
765
+	if (ifindex >= 0)
766
+		n->ifindex = ifindex;
767
+
821768 	return 0;
822769 }
823770 
....@@ -848,16 +795,14 @@
848795 	rcu_assign_pointer(*ins, n);
849796 }
850797 
851
-static struct tc_u_knode *u32_init_knode(struct tcf_proto *tp,
798
+static struct tc_u_knode *u32_init_knode(struct net *net, struct tcf_proto *tp,
852799 					 struct tc_u_knode *n)
853800 {
854801 	struct tc_u_hnode *ht = rtnl_dereference(n->ht_down);
855802 	struct tc_u32_sel *s = &n->sel;
856803 	struct tc_u_knode *new;
857804 
858
-	new = kzalloc(sizeof(*n) + s->nkeys*sizeof(struct tc_u32_key),
859
-		      GFP_KERNEL);
860
-
805
+	new = kzalloc(struct_size(new, sel.keys, s->nkeys), GFP_KERNEL);
861806 	if (!new)
862807 		return NULL;
863808 
....@@ -865,17 +810,10 @@
865810 	new->handle = n->handle;
866811 	RCU_INIT_POINTER(new->ht_up, n->ht_up);
867812 
868
-#ifdef CONFIG_NET_CLS_IND
869813 	new->ifindex = n->ifindex;
870
-#endif
871814 	new->fshift = n->fshift;
872
-	new->res = n->res;
873815 	new->flags = n->flags;
874816 	RCU_INIT_POINTER(new->ht_down, ht);
875
-
876
-	/* bump reference count as long as we hold pointer to structure */
877
-	if (ht)
878
-		ht->refcnt++;
879817 
880818 #ifdef CONFIG_CLS_U32_PERF
881819 	/* Statistics may be incremented by readers during update
....@@ -891,20 +829,23 @@
891829 	/* Similarly success statistics must be moved as pointers */
892830 	new->pcpu_success = n->pcpu_success;
893831 #endif
894
-	new->tp = tp;
895
-	memcpy(&new->sel, s, sizeof(*s) + s->nkeys*sizeof(struct tc_u32_key));
832
+	memcpy(&new->sel, s, struct_size(s, keys, s->nkeys));
896833 
897
-	if (tcf_exts_init(&new->exts, TCA_U32_ACT, TCA_U32_POLICE)) {
834
+	if (tcf_exts_init(&new->exts, net, TCA_U32_ACT, TCA_U32_POLICE)) {
898835 		kfree(new);
899836 		return NULL;
900837 	}
838
+
839
+	/* bump reference count as long as we hold pointer to structure */
840
+	if (ht)
841
+		ht->refcnt++;
901842 
902843 	return new;
903844 }
904845 
905846 static int u32_change(struct net *net, struct sk_buff *in_skb,
906847 		      struct tcf_proto *tp, unsigned long base, u32 handle,
907
-		      struct nlattr **tca, void **arg, bool ovr,
848
+		      struct nlattr **tca, void **arg, bool ovr, bool rtnl_held,
908849 		      struct netlink_ext_ack *extack)
909850 {
910851 	struct tc_u_common *tp_c = tp->data;
....@@ -916,9 +857,6 @@
916857 	u32 htid, flags = 0;
917858 	size_t sel_size;
918859 	int err;
919
-#ifdef CONFIG_CLS_U32_PERF
920
-	size_t size;
921
-#endif
922860 
923861 	if (!opt) {
924862 		if (handle) {
....@@ -929,7 +867,8 @@
929867 		}
930868 	}
931869 
932
-	err = nla_parse_nested(tb, TCA_U32_MAX, opt, u32_policy, extack);
870
+	err = nla_parse_nested_deprecated(tb, TCA_U32_MAX, opt, u32_policy,
871
+					  extack);
933872 	if (err < 0)
934873 		return err;
935874 
....@@ -956,22 +895,21 @@
956895 			return -EINVAL;
957896 		}
958897 
959
-		new = u32_init_knode(tp, n);
898
+		new = u32_init_knode(net, tp, n);
960899 		if (!new)
961900 			return -ENOMEM;
962901 
963
-		err = u32_set_parms(net, tp, base,
964
-				    rtnl_dereference(n->ht_up), new, tb,
902
+		err = u32_set_parms(net, tp, base, new, tb,
965903 				    tca[TCA_RATE], ovr, extack);
966904 
967905 		if (err) {
968
-			u32_destroy_key(tp, new, false);
906
+			__u32_destroy_key(new);
969907 			return err;
970908 		}
971909 
972910 		err = u32_replace_hw_knode(tp, new, flags, extack);
973911 		if (err) {
974
-			u32_destroy_key(tp, new, false);
912
+			__u32_destroy_key(new);
975913 			return err;
976914 		}
977915 
....@@ -988,7 +926,11 @@
988926 	if (tb[TCA_U32_DIVISOR]) {
989927 		unsigned int divisor = nla_get_u32(tb[TCA_U32_DIVISOR]);
990928 
991
-		if (--divisor > 0x100) {
929
+		if (!is_power_of_2(divisor)) {
930
+			NL_SET_ERR_MSG_MOD(extack, "Divisor is not a power of 2");
931
+			return -EINVAL;
932
+		}
933
+		if (divisor-- > 0x100) {
992934 			NL_SET_ERR_MSG_MOD(extack, "Exceeded maximum 256 hash buckets");
993935 			return -EINVAL;
994936 		}
....@@ -996,7 +938,7 @@
996938 			NL_SET_ERR_MSG_MOD(extack, "Divisor can only be used on a hash table");
997939 			return -EINVAL;
998940 		}
999
-		ht = kzalloc(sizeof(*ht) + divisor*sizeof(void *), GFP_KERNEL);
941
+		ht = kzalloc(struct_size(ht, ht, divisor + 1), GFP_KERNEL);
1000942 		if (ht == NULL)
1001943 			return -ENOBUFS;
1002944 		if (handle == 0) {
....@@ -1013,7 +955,6 @@
1013955 				return err;
1014956 			}
1015957 		}
1016
-		ht->tp_c = tp_c;
1017958 		ht->refcnt = 1;
1018959 		ht->divisor = divisor;
1019960 		ht->handle = handle;
....@@ -1057,18 +998,62 @@
1057998 		return -EINVAL;
1058999 	}
10591000 
1001
+	/* At this point, we need to derive the new handle that will be used to
1002
+	 * uniquely map the identity of this table match entry. The
1003
+	 * identity of the entry that we need to construct is 32 bits made of:
1004
+	 *     htid(12b):bucketid(8b):node/entryid(12b)
1005
+	 *
1006
+	 * At this point _we have the table(ht)_ in which we will insert this
1007
+	 * entry. We carry the table's id in variable "htid".
1008
+	 * Note that earlier code picked the ht selection either by a) the user
1009
+	 * providing the htid specified via TCA_U32_HASH attribute or b) when
1010
+	 * no such attribute is passed then the root ht, is default to at ID
1011
+	 * 0x[800][00][000]. Rule: the root table has a single bucket with ID 0.
1012
+	 * If OTOH the user passed us the htid, they may also pass a bucketid of
1013
+	 * choice. 0 is fine. For example a user htid is 0x[600][01][000] it is
1014
+	 * indicating hash bucketid of 1. Rule: the entry/node ID _cannot_ be
1015
+	 * passed via the htid, so even if it was non-zero it will be ignored.
1016
+	 *
1017
+	 * We may also have a handle, if the user passed one. The handle also
1018
+	 * carries the same addressing of htid(12b):bucketid(8b):node/entryid(12b).
1019
+	 * Rule: the bucketid on the handle is ignored even if one was passed;
1020
+	 * rather the value on "htid" is always assumed to be the bucketid.
1021
+	 */
10601022 	if (handle) {
1023
+		/* Rule: The htid from handle and tableid from htid must match */
10611024 		if (TC_U32_HTID(handle) && TC_U32_HTID(handle ^ htid)) {
10621025 			NL_SET_ERR_MSG_MOD(extack, "Handle specified hash table address mismatch");
10631026 			return -EINVAL;
10641027 		}
1065
-		handle = htid | TC_U32_NODE(handle);
1066
-		err = idr_alloc_u32(&ht->handle_idr, NULL, &handle, handle,
1067
-				    GFP_KERNEL);
1068
-		if (err)
1069
-			return err;
1070
-	} else
1028
+		/* Ok, so far we have a valid htid(12b):bucketid(8b) but we
1029
+		 * need to finalize the table entry identification with the last
1030
+		 * part - the node/entryid(12b)). Rule: Nodeid _cannot be 0_ for
1031
+		 * entries. Rule: nodeid of 0 is reserved only for tables(see
1032
+		 * earlier code which processes TC_U32_DIVISOR attribute).
1033
+		 * Rule: The nodeid can only be derived from the handle (and not
1034
+		 * htid).
1035
+		 * Rule: if the handle specified zero for the node id example
1036
+		 * 0x60000000, then pick a new nodeid from the pool of IDs
1037
+		 * this hash table has been allocating from.
1038
+		 * If OTOH it is specified (i.e for example the user passed a
1039
+		 * handle such as 0x60000123), then we use it generate our final
1040
+		 * handle which is used to uniquely identify the match entry.
1041
+		 */
1042
+		if (!TC_U32_NODE(handle)) {
1043
+			handle = gen_new_kid(ht, htid);
1044
+		} else {
1045
+			handle = htid | TC_U32_NODE(handle);
1046
+			err = idr_alloc_u32(&ht->handle_idr, NULL, &handle,
1047
+					    handle, GFP_KERNEL);
1048
+			if (err)
1049
+				return err;
1050
+		}
1051
+	} else {
1052
+		/* The user did not give us a handle; lets just generate one
1053
+		 * from the table's pool of nodeids.
1054
+		 */
10711055 		handle = gen_new_kid(ht, htid);
1056
+	}
10721057 
10731058 	if (tb[TCA_U32_SEL] == NULL) {
10741059 		NL_SET_ERR_MSG_MOD(extack, "Selector not specified");
....@@ -1083,15 +1068,15 @@
10831068 		goto erridr;
10841069 	}
10851070 
1086
-	n = kzalloc(offsetof(typeof(*n), sel) + sel_size, GFP_KERNEL);
1071
+	n = kzalloc(struct_size(n, sel.keys, s->nkeys), GFP_KERNEL);
10871072 	if (n == NULL) {
10881073 		err = -ENOBUFS;
10891074 		goto erridr;
10901075 	}
10911076 
10921077 #ifdef CONFIG_CLS_U32_PERF
1093
-	size = sizeof(struct tc_u32_pcnt) + s->nkeys * sizeof(u64);
1094
-	n->pf = __alloc_percpu(size, __alignof__(struct tc_u32_pcnt));
1078
+	n->pf = __alloc_percpu(struct_size(n->pf, kcnts, s->nkeys),
1079
+			       __alignof__(struct tc_u32_pcnt));
10951080 	if (!n->pf) {
10961081 		err = -ENOBUFS;
10971082 		goto errfree;
....@@ -1103,9 +1088,8 @@
11031088 	n->handle = handle;
11041089 	n->fshift = s->hmask ? ffs(ntohl(s->hmask)) - 1 : 0;
11051090 	n->flags = flags;
1106
-	n->tp = tp;
11071091 
1108
-	err = tcf_exts_init(&n->exts, TCA_U32_ACT, TCA_U32_POLICE);
1092
+	err = tcf_exts_init(&n->exts, net, TCA_U32_ACT, TCA_U32_POLICE);
11091093 	if (err < 0)
11101094 		goto errout;
11111095 
....@@ -1125,7 +1109,7 @@
11251109 	}
11261110 #endif
11271111 
1128
-	err = u32_set_parms(net, tp, base, ht, n, tb, tca[TCA_RATE], ovr,
1112
+	err = u32_set_parms(net, tp, base, n, tb, tca[TCA_RATE], ovr,
11291113 			    extack);
11301114 	if (err == 0) {
11311115 		struct tc_u_knode __rcu **ins;
....@@ -1146,6 +1130,7 @@
11461130 
11471131 		RCU_INIT_POINTER(n->next, pins);
11481132 		rcu_assign_pointer(*ins, n);
1133
+		tp_c->knodes++;
11491134 		*arg = n;
11501135 		return 0;
11511136 	}
....@@ -1167,7 +1152,8 @@
11671152 	return err;
11681153 }
11691154 
1170
-static void u32_walk(struct tcf_proto *tp, struct tcf_walker *arg)
1155
+static void u32_walk(struct tcf_proto *tp, struct tcf_walker *arg,
1156
+		     bool rtnl_held)
11711157 {
11721158 	struct tc_u_common *tp_c = tp->data;
11731159 	struct tc_u_hnode *ht;
....@@ -1208,7 +1194,7 @@
12081194 }
12091195 
12101196 static int u32_reoffload_hnode(struct tcf_proto *tp, struct tc_u_hnode *ht,
1211
-			       bool add, tc_setup_cb_t *cb, void *cb_priv,
1197
+			       bool add, flow_setup_cb_t *cb, void *cb_priv,
12121198 			       struct netlink_ext_ack *extack)
12131199 {
12141200 	struct tc_cls_u32_offload cls_u32 = {};
....@@ -1228,7 +1214,7 @@
12281214 }
12291215 
12301216 static int u32_reoffload_knode(struct tcf_proto *tp, struct tc_u_knode *n,
1231
-			       bool add, tc_setup_cb_t *cb, void *cb_priv,
1217
+			       bool add, flow_setup_cb_t *cb, void *cb_priv,
12321218 			       struct netlink_ext_ack *extack)
12331219 {
12341220 	struct tc_u_hnode *ht = rtnl_dereference(n->ht_down);
....@@ -1251,24 +1237,22 @@
12511237 		cls_u32.knode.mask = 0;
12521238 #endif
12531239 		cls_u32.knode.sel = &n->sel;
1240
+		cls_u32.knode.res = &n->res;
12541241 		cls_u32.knode.exts = &n->exts;
12551242 		if (n->ht_down)
12561243 			cls_u32.knode.link_handle = ht->handle;
12571244 	}
12581245 
1259
-	err = cb(TC_SETUP_CLSU32, &cls_u32, cb_priv);
1260
-	if (err) {
1261
-		if (add && tc_skip_sw(n->flags))
1262
-			return err;
1263
-		return 0;
1264
-	}
1265
-
1266
-	tc_cls_offload_cnt_update(block, &n->in_hw_count, &n->flags, add);
1246
+	err = tc_setup_cb_reoffload(block, tp, add, cb, TC_SETUP_CLSU32,
1247
+				    &cls_u32, cb_priv, &n->flags,
1248
+				    &n->in_hw_count);
1249
+	if (err)
1250
+		return err;
12671251 
12681252 	return 0;
12691253 }
12701254 
1271
-static int u32_reoffload(struct tcf_proto *tp, bool add, tc_setup_cb_t *cb,
1255
+static int u32_reoffload(struct tcf_proto *tp, bool add, flow_setup_cb_t *cb,
12721256 			 void *cb_priv, struct netlink_ext_ack *extack)
12731257 {
12741258 	struct tc_u_common *tp_c = tp->data;
....@@ -1329,7 +1313,7 @@
13291313 }
13301314 
13311315 static int u32_dump(struct net *net, struct tcf_proto *tp, void *fh,
1332
-		    struct sk_buff *skb, struct tcmsg *t)
1316
+		    struct sk_buff *skb, struct tcmsg *t, bool rtnl_held)
13331317 {
13341318 	struct tc_u_knode *n = fh;
13351319 	struct tc_u_hnode *ht_up, *ht_down;
....@@ -1340,7 +1324,7 @@
13401324 
13411325 	t->tcm_handle = n->handle;
13421326 
1343
-	nest = nla_nest_start(skb, TCA_OPTIONS);
1327
+	nest = nla_nest_start_noflag(skb, TCA_OPTIONS);
13441328 	if (nest == NULL)
13451329 		goto nla_put_failure;
13461330 
....@@ -1356,8 +1340,7 @@
13561340 		int cpu;
13571341 #endif
13581342 
1359
-		if (nla_put(skb, TCA_U32_SEL,
1360
-			    sizeof(n->sel) + n->sel.nkeys*sizeof(struct tc_u32_key),
1343
+		if (nla_put(skb, TCA_U32_SEL, struct_size(&n->sel, keys, n->sel.nkeys),
13611344 			    &n->sel))
13621345 			goto nla_put_failure;
13631346 
....@@ -1400,18 +1383,14 @@
14001383 		if (tcf_exts_dump(skb, &n->exts) < 0)
14011384 			goto nla_put_failure;
14021385 
1403
-#ifdef CONFIG_NET_CLS_IND
14041386 		if (n->ifindex) {
14051387 			struct net_device *dev;
14061388 			dev = __dev_get_by_index(net, n->ifindex);
14071389 			if (dev && nla_put_string(skb, TCA_U32_INDEV, dev->name))
14081390 				goto nla_put_failure;
14091391 		}
1410
-#endif
14111392 #ifdef CONFIG_CLS_U32_PERF
1412
-		gpf = kzalloc(sizeof(struct tc_u32_pcnt) +
1413
-			      n->sel.nkeys * sizeof(u64),
1414
-			      GFP_KERNEL);
1393
+		gpf = kzalloc(struct_size(gpf, kcnts, n->sel.nkeys), GFP_KERNEL);
14151394 		if (!gpf)
14161395 			goto nla_put_failure;
14171396 
....@@ -1425,9 +1404,7 @@
14251404 				gpf->kcnts[i] += pf->kcnts[i];
14261405 		}
14271406 
1428
-		if (nla_put_64bit(skb, TCA_U32_PCNT,
1429
-				  sizeof(struct tc_u32_pcnt) +
1430
-				  n->sel.nkeys * sizeof(u64),
1407
+		if (nla_put_64bit(skb, TCA_U32_PCNT, struct_size(gpf, kcnts, n->sel.nkeys),
14311408 				  gpf, TCA_U32_PAD)) {
14321409 			kfree(gpf);
14331410 			goto nla_put_failure;
....@@ -1471,9 +1448,7 @@
14711448 #ifdef CONFIG_CLS_U32_PERF
14721449 	pr_info("    Performance counters on\n");
14731450 #endif
1474
-#ifdef CONFIG_NET_CLS_IND
14751451 	pr_info("    input device check on\n");
1476
-#endif
14771452 #ifdef CONFIG_NET_CLS_ACT
14781453 	pr_info("    Actions configured\n");
14791454 #endif