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

1 files changed, 150 insertions, 0 deletions

diff --git a/tools/testing/radix-tree/benchmark.c b/tools/testing/radix-tree/benchmark.c
new file mode 100644
index 000000000..523c79f22
--- /dev/null
+++ b/tools/testing/radix-tree/benchmark.c
@@ -0,0 +1,150 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * benchmark.c:
+ * Author: Konstantin Khlebnikov <koct9i@gmail.com>
+ */
+#include <linux/radix-tree.h>
+#include <linux/slab.h>
+#include <linux/errno.h>
+#include <time.h>
+#include "test.h"
+
+#define NSEC_PER_SEC	1000000000L
+
+static long long benchmark_iter(struct radix_tree_root *root, bool tagged)
+{
+	volatile unsigned long sink = 0;
+	struct radix_tree_iter iter;
+	struct timespec start, finish;
+	long long nsec;
+	int l, loops = 1;
+	void **slot;
+
+#ifdef BENCHMARK
+again:
+#endif
+	clock_gettime(CLOCK_MONOTONIC, &start);
+	for (l = 0; l < loops; l++) {
+		if (tagged) {
+			radix_tree_for_each_tagged(slot, root, &iter, 0, 0)
+				sink ^= (unsigned long)slot;
+		} else {
+			radix_tree_for_each_slot(slot, root, &iter, 0)
+				sink ^= (unsigned long)slot;
+		}
+	}
+	clock_gettime(CLOCK_MONOTONIC, &finish);
+
+	nsec = (finish.tv_sec - start.tv_sec) * NSEC_PER_SEC +
+	       (finish.tv_nsec - start.tv_nsec);
+
+#ifdef BENCHMARK
+	if (loops == 1 && nsec * 5 < NSEC_PER_SEC) {
+		loops = NSEC_PER_SEC / nsec / 4 + 1;
+		goto again;
+	}
+#endif
+
+	nsec /= loops;
+	return nsec;
+}
+
+static void benchmark_insert(struct radix_tree_root *root,
+			     unsigned long size, unsigned long step)
+{
+	struct timespec start, finish;
+	unsigned long index;
+	long long nsec;
+
+	clock_gettime(CLOCK_MONOTONIC, &start);
+
+	for (index = 0 ; index < size ; index += step)
+		item_insert(root, index);
+
+	clock_gettime(CLOCK_MONOTONIC, &finish);
+
+	nsec = (finish.tv_sec - start.tv_sec) * NSEC_PER_SEC +
+	       (finish.tv_nsec - start.tv_nsec);
+
+	printv(2, "Size: %8ld, step: %8ld, insertion: %15lld ns\n",
+		size, step, nsec);
+}
+
+static void benchmark_tagging(struct radix_tree_root *root,
+			     unsigned long size, unsigned long step)
+{
+	struct timespec start, finish;
+	unsigned long index;
+	long long nsec;
+
+	clock_gettime(CLOCK_MONOTONIC, &start);
+
+	for (index = 0 ; index < size ; index += step)
+		radix_tree_tag_set(root, index, 0);
+
+	clock_gettime(CLOCK_MONOTONIC, &finish);
+
+	nsec = (finish.tv_sec - start.tv_sec) * NSEC_PER_SEC +
+	       (finish.tv_nsec - start.tv_nsec);
+
+	printv(2, "Size: %8ld, step: %8ld, tagging: %17lld ns\n",
+		size, step, nsec);
+}
+
+static void benchmark_delete(struct radix_tree_root *root,
+			     unsigned long size, unsigned long step)
+{
+	struct timespec start, finish;
+	unsigned long index;
+	long long nsec;
+
+	clock_gettime(CLOCK_MONOTONIC, &start);
+
+	for (index = 0 ; index < size ; index += step)
+		item_delete(root, index);
+
+	clock_gettime(CLOCK_MONOTONIC, &finish);
+
+	nsec = (finish.tv_sec - start.tv_sec) * NSEC_PER_SEC +
+	       (finish.tv_nsec - start.tv_nsec);
+
+	printv(2, "Size: %8ld, step: %8ld, deletion: %16lld ns\n",
+		size, step, nsec);
+}
+
+static void benchmark_size(unsigned long size, unsigned long step)
+{
+	RADIX_TREE(tree, GFP_KERNEL);
+	long long normal, tagged;
+
+	benchmark_insert(&tree, size, step);
+	benchmark_tagging(&tree, size, step);
+
+	tagged = benchmark_iter(&tree, true);
+	normal = benchmark_iter(&tree, false);
+
+	printv(2, "Size: %8ld, step: %8ld, tagged iteration: %8lld ns\n",
+		size, step, tagged);
+	printv(2, "Size: %8ld, step: %8ld, normal iteration: %8lld ns\n",
+		size, step, normal);
+
+	benchmark_delete(&tree, size, step);
+
+	item_kill_tree(&tree);
+	rcu_barrier();
+}
+
+void benchmark(void)
+{
+	unsigned long size[] = {1 << 10, 1 << 20, 0};
+	unsigned long step[] = {1, 2, 7, 15, 63, 64, 65,
+				128, 256, 512, 12345, 0};
+	int c, s;
+
+	printv(1, "starting benchmarks\n");
+	printv(1, "RADIX_TREE_MAP_SHIFT = %d\n", RADIX_TREE_MAP_SHIFT);
+
+	for (c = 0; size[c]; c++)
+		for (s = 0; step[s]; s++)
+			benchmark_size(size[c], step[s]);
+}