sched/cache: Keep nr_pref_llc_running in the runnable domain, to fix LLC mis-scheduling buggrafted

alb_break_llc() decides whether to break LLC preference during active
load balance. It does so by testing that every runnable fair task on the
source rq prefers its LLC:

	env->src_rq->nr_pref_llc_running == env->src_rq->cfs.h_nr_runnable

But the two counters cover different sets. nr_pref_llc_running is updated
in account_llc_enqueue()/account_llc_dequeue(), next to cfs_rq->nr_queued,
so it follows queued tasks. h_nr_runnable is updated in set_delayed()/
clear_delayed() and drops delay-dequeued tasks.

So under DELAY_DEQUEUE, a preferring task that goes to sleep stays counted
in nr_pref_llc_running while h_nr_runnable falls. The equality then breaks,
alb_break_llc() returns false, and active balance is free to pull a task
off its preferred LLC. Active balance only moves runnable tasks, and this
is the only LLC check it consults: once the stopper runs, LBF_ACTIVE_LB
skips the per-task test in can_migrate_task(). The runnable set is the one
we want.

Fix it on the counter side. A task should be counted in
nr_pref_llc_running exactly while it is both queued on its preferred LLC
(pref_llc_queued) and runnable (!sched_delayed). Define that membership
once in task_pref_llc_runnable(), and adjust the counter only through
pref_llc_running_inc()/pref_llc_running_dec() from the four sites that
change either input: account_llc_enqueue(), account_llc_dequeue(),
set_delayed() and clear_delayed(). Gating every update on the same
predicate keeps the delay, wake and dequeue paths from double-counting
or underflowing; see the comments at those sites for the ordering.

nr_llc_running and sd->llc_counts are not touched and stay on queued
semantics.

Fixes: 714059f79ff0 ("sched/cache: Handle moving single tasks to/from their preferred LLC")
Closes: https://lore.kernel.org/lkml/20260827135000.735138-1-zhanxusheng@xiaomi.com/
Reported-by: Zhan Xusheng <zhanxusheng@xiaomi.com>
Suggested-by: Chen Yu <yu.c.chen@intel.com>
Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Kayra Cizmeci <kayracizmeci@gmail.com>
Cc: <stable@kernel.org> # v7.2.x
Link: https://patch.msgid.link/06af61afedac32e6477f57feb4d658f6c411c3af.1790035273.git.tim.c.chen@linux.intel.com

4 files changed, 143 insertions, 0 deletions

diff --git a/tools/testing/rbtree/Makefile b/tools/testing/rbtree/Makefile
new file mode 100644
index 000000000..d7bbae2af
--- /dev/null
+++ b/tools/testing/rbtree/Makefile
@@ -0,0 +1,33 @@
+# SPDX-License-Identifier: GPL-2.0
+
+.PHONY: clean
+
+TARGETS = rbtree_test interval_tree_test
+OFILES = $(SHARED_OFILES) rbtree-shim.o interval_tree-shim.o maple-shim.o
+DEPS = ../../../include/linux/rbtree.h \
+	../../../include/linux/rbtree_types.h \
+	../../../include/linux/rbtree_augmented.h \
+	../../../include/linux/interval_tree.h \
+	../../../include/linux/interval_tree_generic.h \
+	../../../lib/rbtree.c \
+	../../../lib/interval_tree.c
+
+targets: $(TARGETS)
+
+include ../shared/shared.mk
+
+ifeq ($(DEBUG), 1)
+	CFLAGS += -g
+endif
+
+$(TARGETS):	$(OFILES)
+
+rbtree-shim.o: $(DEPS)
+rbtree_test.o:  ../../../lib/rbtree_test.c
+interval_tree-shim.o: $(DEPS)
+interval_tree-shim.o: CFLAGS += -DCONFIG_INTERVAL_TREE_SPAN_ITER
+interval_tree_test.o: 	../../../lib/interval_tree_test.c
+interval_tree_test.o: CFLAGS += -DCONFIG_INTERVAL_TREE_SPAN_ITER
+
+clean:
+	$(RM) $(TARGETS) *.o radix-tree.c idr.c generated/*
diff --git a/tools/testing/rbtree/interval_tree_test.c b/tools/testing/rbtree/interval_tree_test.c
new file mode 100644
index 000000000..49bc5b534
--- /dev/null
+++ b/tools/testing/rbtree/interval_tree_test.c
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * interval_tree.c: Userspace Interval Tree test-suite
+ * Copyright (c) 2025 Wei Yang <richard.weiyang@gmail.com>
+ */
+#include <linux/math64.h>
+#include <linux/kern_levels.h>
+#include "shared.h"
+#include "maple-shared.h"
+
+#include "../../../lib/interval_tree_test.c"
+
+int usage(void)
+{
+	fprintf(stderr, "Userland interval tree test cases\n");
+	fprintf(stderr, "  -n: Number of nodes in the interval tree\n");
+	fprintf(stderr, "  -p: Number of iterations modifying the tree\n");
+	fprintf(stderr, "  -q: Number of searches to the interval tree\n");
+	fprintf(stderr, "  -s: Number of iterations searching the tree\n");
+	fprintf(stderr, "  -a: Searches will iterate all nodes in the tree\n");
+	fprintf(stderr, "  -m: Largest value for the interval's endpoint\n");
+	fprintf(stderr, "  -r: Random seed\n");
+	exit(-1);
+}
+
+void interval_tree_tests(void)
+{
+	interval_tree_test_init();
+	interval_tree_test_exit();
+}
+
+int main(int argc, char **argv)
+{
+	int opt;
+
+	while ((opt = getopt(argc, argv, "n:p:q:s:am:r:")) != -1) {
+		if (opt == 'n')
+			nnodes = strtoul(optarg, NULL, 0);
+		else if (opt == 'p')
+			perf_loops = strtoul(optarg, NULL, 0);
+		else if (opt == 'q')
+			nsearches = strtoul(optarg, NULL, 0);
+		else if (opt == 's')
+			search_loops = strtoul(optarg, NULL, 0);
+		else if (opt == 'a')
+			search_all = true;
+		else if (opt == 'm')
+			max_endpoint = strtoul(optarg, NULL, 0);
+		else if (opt == 'r')
+			seed = strtoul(optarg, NULL, 0);
+		else
+			usage();
+	}
+
+	maple_tree_init();
+	interval_tree_tests();
+	return 0;
+}
diff --git a/tools/testing/rbtree/rbtree_test.c b/tools/testing/rbtree/rbtree_test.c
new file mode 100644
index 000000000..585c970f6
--- /dev/null
+++ b/tools/testing/rbtree/rbtree_test.c
@@ -0,0 +1,48 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * rbtree_test.c: Userspace Red Black Tree test-suite
+ * Copyright (c) 2025 Wei Yang <richard.weiyang@gmail.com>
+ */
+#include <linux/init.h>
+#include <linux/math64.h>
+#include <linux/kern_levels.h>
+#include "shared.h"
+
+#include "../../../lib/rbtree_test.c"
+
+int usage(void)
+{
+	fprintf(stderr, "Userland rbtree test cases\n");
+	fprintf(stderr, "  -n: Number of nodes in the rb-tree\n");
+	fprintf(stderr, "  -p: Number of iterations modifying the rb-tree\n");
+	fprintf(stderr, "  -c: Number of iterations modifying and verifying the rb-tree\n");
+	fprintf(stderr, "  -r: Random seed\n");
+	exit(-1);
+}
+
+void rbtree_tests(void)
+{
+	rbtree_test_init();
+	rbtree_test_exit();
+}
+
+int main(int argc, char **argv)
+{
+	int opt;
+
+	while ((opt = getopt(argc, argv, "n:p:c:r:")) != -1) {
+		if (opt == 'n')
+			nnodes = strtoul(optarg, NULL, 0);
+		else if (opt == 'p')
+			perf_loops = strtoul(optarg, NULL, 0);
+		else if (opt == 'c')
+			check_loops = strtoul(optarg, NULL, 0);
+		else if (opt == 'r')
+			seed = strtoul(optarg, NULL, 0);
+		else
+			usage();
+	}
+
+	rbtree_tests();
+	return 0;
+}
diff --git a/tools/testing/rbtree/test.h b/tools/testing/rbtree/test.h
new file mode 100644
index 000000000..f1f1b545b
--- /dev/null
+++ b/tools/testing/rbtree/test.h
@@ -0,0 +1,4 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+
+void rbtree_tests(void);
+void interval_tree_tests(void);