From 0d6526f82c3cdefcca47f73f5fc08dc6f335eac6 Mon Sep 17 00:00:00 2001
From: Tim Chen <tim.c.chen@linux.intel.com>
Date: Mon, 21 Sep 2026 17:37:22 -0700
Subject: sched/cache: Keep nr_pref_llc_running in the runnable domain, to fix
 LLC mis-scheduling bug

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
---
 scripts/gfp-translate | 118 ++++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 118 insertions(+)
 create mode 100755 scripts/gfp-translate

(limited to 'scripts/gfp-translate')

diff --git a/scripts/gfp-translate b/scripts/gfp-translate
new file mode 100755
index 000000000..8385ae0d5
--- /dev/null
+++ b/scripts/gfp-translate
@@ -0,0 +1,118 @@
+#!/bin/bash
+# SPDX-License-Identifier: GPL-2.0-only
+# Translate the bits making up a GFP mask
+# (c) 2009, Mel Gorman <mel@csn.ul.ie>
+SOURCE=
+GFPMASK=none
+
+# Helper function to report failures and exit
+die() {
+	echo ERROR: $@
+	if [ "$TMPFILE" != "" ]; then
+		rm -f $TMPFILE
+	fi
+	exit -1
+}
+
+usage() {
+	echo "usage: gfp-translate [-h] [ --source DIRECTORY ] gfpmask"
+	exit 0
+}
+
+# Parse command-line arguments
+while [ $# -gt 0 ]; do
+	case $1 in
+		--source)
+			SOURCE=$2
+			shift 2
+			;;
+		-h)
+			usage
+			;;
+		--help)
+			usage
+			;;
+		*)
+			GFPMASK=$1
+			shift
+			;;
+	esac
+done
+
+# Guess the kernel source directory if it's not set. Preference is in order of
+# o current directory
+# o /usr/src/linux
+if [ "$SOURCE" = "" ]; then
+	if [ -r "/usr/src/linux/Makefile" ]; then
+		SOURCE=/usr/src/linux
+	fi
+	if [ -r "`pwd`/Makefile" ]; then
+		SOURCE=`pwd`
+	fi
+fi
+
+# Confirm that a source directory exists
+if [ ! -r "$SOURCE/Makefile" ]; then
+	die "Could not locate kernel source directory or it is invalid"
+fi
+
+# Confirm that a GFP mask has been specified
+if [ "$GFPMASK" = "none" ]; then
+	usage
+fi
+
+# Extract GFP flags from the kernel source
+TMPFILE=`mktemp -t gfptranslate-XXXXXX.c` || exit 1
+
+echo Source: $SOURCE
+echo Parsing: $GFPMASK
+
+(
+    cat <<EOF
+#include <stdint.h>
+#include <stdio.h>
+
+// Try to fool compiler.h into not including extra stuff
+#define __ASSEMBLY__	1
+
+#include <generated/autoconf.h>
+#include <linux/gfp_types.h>
+
+static const char *masks[] = {
+EOF
+
+    sed -nEe 's/^[[:space:]]+(___GFP_.*)_BIT,.*$/\1/p' $SOURCE/include/linux/gfp_types.h |
+	while read b; do
+	    cat <<EOF
+#if defined($b) && ($b > 0)
+	[${b}_BIT]	= "$b",
+#endif
+EOF
+	done
+
+    cat <<EOF
+};
+
+int main(int argc, char *argv[])
+{
+	unsigned long long mask = $GFPMASK;
+
+	for (int i = 0; i < sizeof(mask) * 8; i++) {
+		unsigned long long bit = 1ULL << i;
+		if (mask & bit)
+			printf("\t%-25s0x%llx\n",
+			       (i < ___GFP_LAST_BIT && masks[i]) ?
+					masks[i] : "*** INVALID ***",
+			       bit);
+	}
+
+	return 0;
+}
+EOF
+) > $TMPFILE
+
+${CC:-gcc} -Wall -o ${TMPFILE}.bin -I $SOURCE/include $TMPFILE && ${TMPFILE}.bin
+
+rm -f $TMPFILE ${TMPFILE}.bin
+
+exit 0
-- 
cgit v1.3.1