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, 142 insertions, 0 deletions

diff --git a/drivers/input/joystick/joydump.c b/drivers/input/joystick/joydump.c
new file mode 100644
index 000000000..c9cbdcc7f
--- /dev/null
+++ b/drivers/input/joystick/joydump.c
@@ -0,0 +1,142 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ *  Copyright (c) 1996-2001 Vojtech Pavlik
+ */
+
+/*
+ * This is just a very simple driver that can dump the data
+ * out of the joystick port into the syslog ...
+ */
+
+#include <linux/module.h>
+#include <linux/gameport.h>
+#include <linux/kernel.h>
+#include <linux/delay.h>
+#include <linux/slab.h>
+
+#define DRIVER_DESC	"Gameport data dumper module"
+
+MODULE_AUTHOR("Vojtech Pavlik <vojtech@ucw.cz>");
+MODULE_DESCRIPTION(DRIVER_DESC);
+MODULE_LICENSE("GPL");
+
+#define BUF_SIZE 256
+
+struct joydump {
+	unsigned int time;
+	unsigned char data;
+};
+
+static int joydump_connect(struct gameport *gameport, struct gameport_driver *drv)
+{
+	struct joydump *buf;	/* all entries */
+	struct joydump *dump, *prev;	/* one entry each */
+	int axes[4], buttons;
+	int i, j, t, timeout;
+	unsigned long flags;
+	unsigned char u;
+
+	printk(KERN_INFO "joydump: ,------------------ START ----------------.\n");
+	printk(KERN_INFO "joydump: | Dumping: %30s |\n", gameport->phys);
+	printk(KERN_INFO "joydump: | Speed: %28d kHz |\n", gameport->speed);
+
+	if (gameport_open(gameport, drv, GAMEPORT_MODE_RAW)) {
+
+		printk(KERN_INFO "joydump: | Raw mode not available - trying cooked.    |\n");
+
+		if (gameport_open(gameport, drv, GAMEPORT_MODE_COOKED)) {
+
+			printk(KERN_INFO "joydump: | Cooked not available either. Failing.   |\n");
+			printk(KERN_INFO "joydump: `------------------- END -----------------'\n");
+			return -ENODEV;
+		}
+
+		gameport_cooked_read(gameport, axes, &buttons);
+
+		for (i = 0; i < 4; i++)
+			printk(KERN_INFO "joydump: | Axis %d: %4d.                           |\n", i, axes[i]);
+		printk(KERN_INFO "joydump: | Buttons %02x.                             |\n", buttons);
+		printk(KERN_INFO "joydump: `------------------- END -----------------'\n");
+	}
+
+	timeout = gameport_time(gameport, 10000); /* 10 ms */
+
+	buf = kmalloc_objs(struct joydump, BUF_SIZE);
+	if (!buf) {
+		printk(KERN_INFO "joydump: no memory for testing\n");
+		goto jd_end;
+	}
+	dump = buf;
+	t = 0;
+	i = 1;
+
+	local_irq_save(flags);
+
+	u = gameport_read(gameport);
+
+	dump->data = u;
+	dump->time = t;
+	dump++;
+
+	gameport_trigger(gameport);
+
+	while (i < BUF_SIZE && t < timeout) {
+
+		dump->data = gameport_read(gameport);
+
+		if (dump->data ^ u) {
+			u = dump->data;
+			dump->time = t;
+			i++;
+			dump++;
+		}
+		t++;
+	}
+
+	local_irq_restore(flags);
+
+/*
+ * Dump data.
+ */
+
+	t = i;
+	dump = buf;
+	prev = dump;
+
+	printk(KERN_INFO "joydump: >------------------ DATA -----------------<\n");
+	printk(KERN_INFO "joydump: | index: %3d delta: %3d us data: ", 0, 0);
+	for (j = 7; j >= 0; j--)
+		printk("%d", (dump->data >> j) & 1);
+	printk(" |\n");
+	dump++;
+
+	for (i = 1; i < t; i++, dump++, prev++) {
+		printk(KERN_INFO "joydump: | index: %3d delta: %3d us data: ",
+			i, dump->time - prev->time);
+		for (j = 7; j >= 0; j--)
+			printk("%d", (dump->data >> j) & 1);
+		printk(" |\n");
+	}
+	kfree(buf);
+
+jd_end:
+	printk(KERN_INFO "joydump: `------------------- END -----------------'\n");
+
+	return 0;
+}
+
+static void joydump_disconnect(struct gameport *gameport)
+{
+	gameport_close(gameport);
+}
+
+static struct gameport_driver joydump_drv = {
+	.driver		= {
+		.name	= "joydump",
+	},
+	.description	= DRIVER_DESC,
+	.connect	= joydump_connect,
+	.disconnect	= joydump_disconnect,
+};
+
+module_gameport_driver(joydump_drv);