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, 110 insertions, 0 deletions

diff --git a/tools/bootconfig/scripts/ftrace.sh b/tools/bootconfig/scripts/ftrace.sh
new file mode 100644
index 000000000..cc5250c64
--- /dev/null
+++ b/tools/bootconfig/scripts/ftrace.sh
@@ -0,0 +1,110 @@
+#!/bin/sh
+# SPDX-License-Identifier: GPL-2.0-only
+
+clear_trace() { # reset trace output
+    echo > trace
+}
+
+disable_tracing() { # stop trace recording
+    echo 0 > tracing_on
+}
+
+enable_tracing() { # start trace recording
+    echo 1 > tracing_on
+}
+
+reset_tracer() { # reset the current tracer
+    echo nop > current_tracer
+}
+
+reset_trigger_file() {
+    # remove action triggers first
+    grep -H ':on[^:]*(' $@ |
+    while read line; do
+        cmd=`echo $line | cut -f2- -d: | cut -f1 -d"["`
+	file=`echo $line | cut -f1 -d:`
+	echo "!$cmd" >> $file
+    done
+    grep -Hv ^# $@ |
+    while read line; do
+        cmd=`echo $line | cut -f2- -d: | cut -f1 -d"["`
+	file=`echo $line | cut -f1 -d:`
+	echo "!$cmd" > $file
+    done
+}
+
+reset_trigger() { # reset all current setting triggers
+    if [ -d events/synthetic ]; then
+        reset_trigger_file events/synthetic/*/trigger
+    fi
+    reset_trigger_file events/*/*/trigger
+}
+
+reset_events_filter() { # reset all current setting filters
+    grep -v ^none events/*/*/filter |
+    while read line; do
+	echo 0 > `echo $line | cut -f1 -d:`
+    done
+}
+
+reset_ftrace_filter() { # reset all triggers in set_ftrace_filter
+    if [ ! -f set_ftrace_filter ]; then
+      return 0
+    fi
+    echo > set_ftrace_filter
+    grep -v '^#' set_ftrace_filter | while read t; do
+	tr=`echo $t | cut -d: -f2`
+	if [ "$tr" = "" ]; then
+	    continue
+	fi
+	if ! grep -q "$t" set_ftrace_filter; then
+		continue;
+	fi
+	name=`echo $t | cut -d: -f1 | cut -d' ' -f1`
+	if [ $tr = "enable_event" -o $tr = "disable_event" ]; then
+	    tr=`echo $t | cut -d: -f2-4`
+	    limit=`echo $t | cut -d: -f5`
+	else
+	    tr=`echo $t | cut -d: -f2`
+	    limit=`echo $t | cut -d: -f3`
+	fi
+	if [ "$limit" != "unlimited" ]; then
+	    tr="$tr:$limit"
+	fi
+	echo "!$name:$tr" > set_ftrace_filter
+    done
+}
+
+disable_events() {
+    echo 0 > events/enable
+}
+
+clear_synthetic_events() { # reset all current synthetic events
+    grep -v ^# synthetic_events |
+    while read line; do
+        echo "!$line" >> synthetic_events
+    done
+}
+
+initialize_ftrace() { # Reset ftrace to initial-state
+# As the initial state, ftrace will be set to nop tracer,
+# no events, no triggers, no filters, no function filters,
+# no probes, and tracing on.
+    disable_tracing
+    reset_tracer
+    reset_trigger
+    reset_events_filter
+    reset_ftrace_filter
+    disable_events
+    [ -f set_event_pid ] && echo > set_event_pid
+    [ -f set_ftrace_pid ] && echo > set_ftrace_pid
+    [ -f set_ftrace_notrace ] && echo > set_ftrace_notrace
+    [ -f set_graph_function ] && echo | tee set_graph_*
+    [ -f stack_trace_filter ] && echo > stack_trace_filter
+    [ -f kprobe_events ] && echo > kprobe_events
+    [ -f uprobe_events ] && echo > uprobe_events
+    [ -f synthetic_events ] && echo > synthetic_events
+    [ -f snapshot ] && echo 0 > snapshot
+    clear_trace
+    enable_tracing
+}