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, 184 insertions, 0 deletions

diff --git a/scripts/atomic/gen-atomic-instrumented.sh b/scripts/atomic/gen-atomic-instrumented.sh
new file mode 100755
index 000000000..9c1d53f81
--- /dev/null
+++ b/scripts/atomic/gen-atomic-instrumented.sh
@@ -0,0 +1,184 @@
+#!/bin/sh
+# SPDX-License-Identifier: GPL-2.0
+
+ATOMICDIR=$(dirname $0)
+
+. ${ATOMICDIR}/atomic-tbl.sh
+
+#gen_param_check(meta, arg)
+gen_param_check()
+{
+	local meta="$1"; shift
+	local arg="$1"; shift
+	local type="${arg%%:*}"
+	local name="$(gen_param_name "${arg}")"
+	local rw="atomic_write"
+
+	case "${type#c}" in
+	i) return;;
+	esac
+
+	if [ ${type#c} != ${type} ]; then
+		# We don't write to constant parameters.
+		rw="atomic_read"
+	elif [ "${type}" = "p" ] ; then
+		# The "old" argument in try_cmpxchg() gets accessed non-atomically
+		rw="read_write"
+	elif [ "${meta}" != "s" ]; then
+		# An atomic RMW: if this parameter is not a constant, and this atomic is
+		# not just a 's'tore, this parameter is both read from and written to.
+		rw="atomic_read_write"
+	fi
+
+	printf "\tinstrument_${rw}(${name}, sizeof(*${name}));\n"
+}
+
+#gen_params_checks(meta, arg...)
+gen_params_checks()
+{
+	local meta="$1"; shift
+	local order="$1"; shift
+
+	if [ "${order}" = "_release" ]; then
+		printf "\tkcsan_release();\n"
+	elif [ -z "${order}" ] && ! meta_in "$meta" "slv"; then
+		# RMW with return value is fully ordered
+		printf "\tkcsan_mb();\n"
+	fi
+
+	while [ "$#" -gt 0 ]; do
+		gen_param_check "$meta" "$1"
+		shift;
+	done
+}
+
+#gen_proto_order_variant(meta, pfx, name, sfx, order, atomic, int, arg...)
+gen_proto_order_variant()
+{
+	local meta="$1"; shift
+	local pfx="$1"; shift
+	local name="$1"; shift
+	local sfx="$1"; shift
+	local order="$1"; shift
+	local atomic="$1"; shift
+	local int="$1"; shift
+
+	local atomicname="${atomic}_${pfx}${name}${sfx}${order}"
+
+	local ret="$(gen_ret_type "${meta}" "${int}")"
+	local params="$(gen_params "${int}" "${atomic}" "$@")"
+	local checks="$(gen_params_checks "${meta}" "${order}" "$@")"
+	local args="$(gen_args "$@")"
+	local retstmt="$(gen_ret_stmt "${meta}")"
+
+	gen_kerneldoc "" "${meta}" "${pfx}" "${name}" "${sfx}" "${order}" "${atomic}" "${int}" "$@"
+
+cat <<EOF
+static __always_inline ${ret}
+${atomicname}(${params})
+{
+${checks}
+	${retstmt}raw_${atomicname}(${args});
+}
+EOF
+
+	printf "\n"
+}
+
+gen_xchg()
+{
+	local xchg="$1"; shift
+	local order="$1"; shift
+
+	kcsan_barrier=""
+	if [ "${xchg%_local}" = "${xchg}" ]; then
+		case "$order" in
+		_release)	kcsan_barrier="kcsan_release()" ;;
+		"")			kcsan_barrier="kcsan_mb()" ;;
+		esac
+	fi
+
+	if [ "${xchg%${xchg#try_cmpxchg}}" = "try_cmpxchg" ] ; then
+
+cat <<EOF
+#define ${xchg}${order}(ptr, oldp, ...) \\
+({ \\
+	typeof(ptr) __ai_ptr = (ptr); \\
+	typeof(oldp) __ai_oldp = (oldp); \\
+EOF
+[ -n "$kcsan_barrier" ] && printf "\t${kcsan_barrier}; \\\\\n"
+cat <<EOF
+	instrument_atomic_read_write(__ai_ptr, sizeof(*__ai_ptr)); \\
+	instrument_read_write(__ai_oldp, sizeof(*__ai_oldp)); \\
+	raw_${xchg}${order}(__ai_ptr, __ai_oldp, __VA_ARGS__); \\
+})
+EOF
+
+	else
+
+cat <<EOF
+#define ${xchg}${order}(ptr, ...) \\
+({ \\
+	typeof(ptr) __ai_ptr = (ptr); \\
+EOF
+[ -n "$kcsan_barrier" ] && printf "\t${kcsan_barrier}; \\\\\n"
+cat <<EOF
+	instrument_atomic_read_write(__ai_ptr, sizeof(*__ai_ptr)); \\
+	raw_${xchg}${order}(__ai_ptr, __VA_ARGS__); \\
+})
+EOF
+
+	fi
+}
+
+cat << EOF
+// SPDX-License-Identifier: GPL-2.0
+
+// Generated by $0
+// DO NOT MODIFY THIS FILE DIRECTLY
+
+/*
+ * This file provoides atomic operations with explicit instrumentation (e.g.
+ * KASAN, KCSAN), which should be used unless it is necessary to avoid
+ * instrumentation. Where it is necessary to aovid instrumenation, the
+ * raw_atomic*() operations should be used.
+ */
+#ifndef _LINUX_ATOMIC_INSTRUMENTED_H
+#define _LINUX_ATOMIC_INSTRUMENTED_H
+
+#include <linux/build_bug.h>
+#include <linux/compiler.h>
+#include <linux/instrumented.h>
+
+EOF
+
+grep '^[a-z]' "$1" | while read name meta args; do
+	gen_proto "${meta}" "${name}" "atomic" "int" ${args}
+done
+
+grep '^[a-z]' "$1" | while read name meta args; do
+	gen_proto "${meta}" "${name}" "atomic64" "s64" ${args}
+done
+
+grep '^[a-z]' "$1" | while read name meta args; do
+	gen_proto "${meta}" "${name}" "atomic_long" "long" ${args}
+done
+
+
+for xchg in "xchg" "cmpxchg" "cmpxchg64" "cmpxchg128" "try_cmpxchg" "try_cmpxchg64" "try_cmpxchg128"; do
+	for order in "" "_acquire" "_release" "_relaxed"; do
+		gen_xchg "${xchg}" "${order}"
+		printf "\n"
+	done
+done
+
+for xchg in "cmpxchg_local" "cmpxchg64_local" "cmpxchg128_local" "sync_cmpxchg" \
+	    "try_cmpxchg_local" "try_cmpxchg64_local" "try_cmpxchg128_local" "sync_try_cmpxchg"; do
+	gen_xchg "${xchg}" ""
+	printf "\n"
+done
+
+cat <<EOF
+
+#endif /* _LINUX_ATOMIC_INSTRUMENTED_H */
+EOF