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, 78 insertions, 0 deletions

diff --git a/scripts/ld-version.sh b/scripts/ld-version.sh
new file mode 100755
index 000000000..b9513d224
--- /dev/null
+++ b/scripts/ld-version.sh
@@ -0,0 +1,78 @@
+#!/bin/sh
+# SPDX-License-Identifier: GPL-2.0
+#
+# Print the linker name and its version in a 5 or 6-digit form.
+# Also, perform the minimum version check.
+
+set -e
+
+# Convert the version string x.y.z to a canonical 5 or 6-digit form.
+get_canonical_version()
+{
+	IFS=.
+	set -- $1
+
+	# If the 2nd or 3rd field is missing, fill it with a zero.
+	#
+	# The 4th field, if present, is ignored.
+	# This occurs in development snapshots as in 2.35.1.20201116
+	echo $((10000 * $1 + 100 * ${2:-0} + ${3:-0}))
+}
+
+orig_args="$@"
+
+# Get the first line of the --version output.
+IFS='
+'
+set -- $(LC_ALL=C "$@" --version)
+
+# Split the line on spaces.
+IFS=' '
+set -- $1
+
+min_tool_version=$(dirname $0)/min-tool-version.sh
+
+if [ "$1" = GNU -a "$2" = ld ]; then
+	shift $(($# - 1))
+	version=$1
+	min_version=$($min_tool_version binutils)
+	name=BFD
+	disp_name="GNU ld"
+elif [ "$1" = GNU -a "$2" = gold ]; then
+	echo "gold linker is not supported as it is not capable of linking the kernel proper." >&2
+	exit 1
+else
+	while [ $# -gt 1 -a "$1" != "LLD" ]; do
+		shift
+	done
+
+	if [ "$1" = LLD ]; then
+		version=$2
+		min_version=$($min_tool_version llvm)
+		name=LLD
+		disp_name=LLD
+	else
+		echo "$orig_args: unknown linker" >&2
+		exit 1
+	fi
+fi
+
+# There may be something after the version, such as a distribution's package
+# release number (like Fedora's "2.34-4.fc32") or punctuation (like LLD briefly
+# added before the "compatible with GNU linkers" string), so remove everything
+# after just numbers and periods.
+version=${version%%[!0-9.]*}
+
+cversion=$(get_canonical_version $version)
+min_cversion=$(get_canonical_version $min_version)
+
+if [ "$cversion" -lt "$min_cversion" ]; then
+	echo >&2 "***"
+	echo >&2 "*** Linker is too old."
+	echo >&2 "***   Your $disp_name version:    $version"
+	echo >&2 "***   Minimum $disp_name version: $min_version"
+	echo >&2 "***"
+	exit 1
+fi
+
+echo $name $cversion