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

3 files changed, 96 insertions, 0 deletions

diff --git a/tools/testing/selftests/module/Makefile b/tools/testing/selftests/module/Makefile
new file mode 100644
index 000000000..6132d7ddb
--- /dev/null
+++ b/tools/testing/selftests/module/Makefile
@@ -0,0 +1,12 @@
+# SPDX-License-Identifier: GPL-2.0-only
+# Makefile for module loading selftests
+
+# No binaries, but make sure arg-less "make" doesn't trigger "run_tests"
+all:
+
+TEST_PROGS := find_symbol.sh
+
+include ../lib.mk
+
+# Nothing to clean up.
+clean:
diff --git a/tools/testing/selftests/module/config b/tools/testing/selftests/module/config
new file mode 100644
index 000000000..b0c206b1a
--- /dev/null
+++ b/tools/testing/selftests/module/config
@@ -0,0 +1,3 @@
+CONFIG_TEST_RUNTIME=y
+CONFIG_TEST_RUNTIME_MODULE=y
+CONFIG_TEST_KALLSYMS=m
diff --git a/tools/testing/selftests/module/find_symbol.sh b/tools/testing/selftests/module/find_symbol.sh
new file mode 100755
index 000000000..2c56805c9
--- /dev/null
+++ b/tools/testing/selftests/module/find_symbol.sh
@@ -0,0 +1,81 @@
+#!/bin/bash
+# SPDX-License-Identifier: GPL-2.0-or-later OR copyleft-next-0.3.1
+# Copyright (C) 2023 Luis Chamberlain <mcgrof@kernel.org>
+#
+# This is a stress test script for kallsyms through find_symbol()
+
+set -e
+
+# Kselftest framework requirement - SKIP code is 4.
+ksft_skip=4
+
+test_reqs()
+{
+	if ! which modprobe 2> /dev/null > /dev/null; then
+		echo "$0: You need modprobe installed" >&2
+		exit $ksft_skip
+	fi
+
+	if ! which kmod 2> /dev/null > /dev/null; then
+		echo "$0: You need kmod installed" >&2
+		exit $ksft_skip
+	fi
+
+	if ! which perf 2> /dev/null > /dev/null; then
+		echo "$0: You need perf installed" >&2
+		exit $ksft_skip
+	fi
+
+	uid=$(id -u)
+	if [ $uid -ne 0 ]; then
+		echo $msg must be run as root >&2
+		exit $ksft_skip
+	fi
+}
+
+load_mod()
+{
+	local STATS="-e duration_time"
+	STATS="$STATS -e user_time"
+	STATS="$STATS -e system_time"
+	STATS="$STATS -e page-faults"
+	local MOD=$1
+
+	local ARCH="$(uname -m)"
+	case "${ARCH}" in
+	x86_64)
+		perf stat $STATS $MODPROBE $MOD
+		;;
+	*)
+		time $MODPROBE $MOD
+		exit 1
+		;;
+	esac
+}
+
+remove_all()
+{
+	$MODPROBE -r test_kallsyms_b
+	for i in a b c d; do
+		$MODPROBE -r test_kallsyms_$i
+	done
+}
+test_reqs
+
+MODPROBE=$(</proc/sys/kernel/modprobe)
+
+remove_all
+load_mod test_kallsyms_b
+remove_all
+
+# Now pollute the namespace
+$MODPROBE test_kallsyms_c
+load_mod test_kallsyms_b
+
+# Now pollute the namespace with twice the number of symbols than the last time
+remove_all
+$MODPROBE test_kallsyms_c
+$MODPROBE test_kallsyms_d
+load_mod test_kallsyms_b
+
+exit 0