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, 162 insertions, 0 deletions

diff --git a/scripts/objdiff b/scripts/objdiff
new file mode 100755
index 000000000..0685bc3ce
--- /dev/null
+++ b/scripts/objdiff
@@ -0,0 +1,162 @@
+#!/bin/bash
+# SPDX-License-Identifier: GPL-2.0-only
+
+# objdiff - a small script for validating that a commit or series of commits
+# didn't change object code.
+#
+# Copyright 2014, Jason Cooper <jason@lakedaemon.net>
+#
+
+# usage example:
+#
+# $ git checkout COMMIT_A
+# $ <your fancy build command here>
+# $ ./scripts/objdiff record path/to/*.o
+#
+# $ git checkout COMMIT_B
+# $ <your fancy build command here>
+# $ ./scripts/objdiff record path/to/*.o
+#
+# $ ./scripts/objdiff diff COMMIT_A COMMIT_B
+# $
+
+# And to clean up (everything is in .objdiff/*)
+# $ ./scripts/objdiff clean all
+#
+# Note: 'make mrproper' will also remove .objdiff
+
+SRCTREE=$(cd $(git rev-parse --show-toplevel 2>/dev/null); pwd)
+
+if [ -z "$SRCTREE" ]; then
+	echo >&2 "ERROR: Not a git repository."
+	exit 1
+fi
+
+TMPD=$SRCTREE/.objdiff
+
+usage() {
+	echo >&2 "Usage: $0 <command> <args>"
+	echo >&2 "  record    <list of object files or directories>"
+	echo >&2 "  diff      <commitA> <commitB>"
+	echo >&2 "  clean     all | <commit>"
+	exit 1
+}
+
+get_output_dir() {
+	dir=${1%/*}
+
+	if [ "$dir" = "$1" ]; then
+		dir=.
+	fi
+
+	dir=$(cd $dir; pwd)
+
+	echo $TMPD/$CMT${dir#$SRCTREE}
+}
+
+do_objdump() {
+	dir=$(get_output_dir $1)
+	base=${1##*/}
+	stripped=$dir/${base%.o}.stripped
+	dis=$dir/${base%.o}.dis
+
+	[ ! -d "$dir" ] && mkdir -p $dir
+
+	# remove addresses for a cleaner diff
+	# http://dummdida.tumblr.com/post/60924060451/binary-diff-between-libc-from-scientificlinux-and
+	$STRIP -g $1 -R __bug_table -R .note -R .comment -o $stripped
+	$OBJDUMP -D $stripped | sed -e "s/^[[:space:]]\+[0-9a-f]\+//" -e "s:^$stripped:$1:" > $dis
+}
+
+dorecord() {
+	[ $# -eq 0 ] && usage
+
+	FILES="$*"
+
+	CMT="`git rev-parse --short HEAD`"
+
+	STRIP="${CROSS_COMPILE}strip"
+	OBJDUMP="${CROSS_COMPILE}objdump"
+
+	for d in $FILES; do
+		if [ -d "$d" ]; then
+			for f in $(find $d -name '*.o')
+			do
+				do_objdump $f
+			done
+		else
+			do_objdump $d
+		fi
+	done
+}
+
+dodiff() {
+	[ $# -ne 2 ] && [ $# -ne 0 ] && usage
+
+	if [ $# -eq 0 ]; then
+		SRC="`git rev-parse --short HEAD^`"
+		DST="`git rev-parse --short HEAD`"
+	else
+		SRC="`git rev-parse --short $1`"
+		DST="`git rev-parse --short $2`"
+	fi
+
+	DIFF="`which colordiff`"
+
+	if [ ${#DIFF} -eq 0 ] || [ ! -x "$DIFF" ]; then
+		DIFF="`which diff`"
+	fi
+
+	SRCD="$TMPD/$SRC"
+	DSTD="$TMPD/$DST"
+
+	if [ ! -d "$SRCD" ]; then
+		echo >&2 "ERROR: $SRCD doesn't exist"
+		exit 1
+	fi
+
+	if [ ! -d "$DSTD" ]; then
+		echo >&2 "ERROR: $DSTD doesn't exist"
+		exit 1
+	fi
+
+	$DIFF -Nurd $SRCD $DSTD
+}
+
+doclean() {
+	[ $# -eq 0 ] && usage
+	[ $# -gt 1 ] && usage
+
+	if [ "x$1" = "xall" ]; then
+		rm -rf $TMPD/*
+	else
+		CMT="`git rev-parse --short $1`"
+
+		if [ -d "$TMPD/$CMT" ]; then
+			rm -rf $TMPD/$CMT
+		else
+			echo >&2 "$CMT not found"
+		fi
+	fi
+}
+
+[ $# -eq 0 ] &&	usage
+
+case "$1" in
+	record)
+		shift
+		dorecord $*
+		;;
+	diff)
+		shift
+		dodiff $*
+		;;
+	clean)
+		shift
+		doclean $*
+		;;
+	*)
+		echo >&2 "Unrecognized command '$1'"
+		exit 1
+		;;
+esac