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, 209 insertions, 0 deletions

diff --git a/drivers/input/serio/ambakmi.c b/drivers/input/serio/ambakmi.c
new file mode 100644
index 000000000..046b8f388
--- /dev/null
+++ b/drivers/input/serio/ambakmi.c
@@ -0,0 +1,209 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ *  linux/drivers/input/serio/ambakmi.c
+ *
+ *  Copyright (C) 2000-2003 Deep Blue Solutions Ltd.
+ *  Copyright (C) 2002 Russell King.
+ */
+#include <linux/module.h>
+#include <linux/serio.h>
+#include <linux/errno.h>
+#include <linux/interrupt.h>
+#include <linux/ioport.h>
+#include <linux/device.h>
+#include <linux/delay.h>
+#include <linux/slab.h>
+#include <linux/err.h>
+#include <linux/amba/bus.h>
+#include <linux/amba/kmi.h>
+#include <linux/clk.h>
+
+#include <asm/io.h>
+#include <asm/irq.h>
+
+#define KMI_BASE	(kmi->base)
+
+struct amba_kmi_port {
+	struct serio		*io;
+	struct clk		*clk;
+	void __iomem		*base;
+	unsigned int		irq;
+	unsigned int		divisor;
+	unsigned int		open;
+};
+
+static irqreturn_t amba_kmi_int(int irq, void *dev_id)
+{
+	struct amba_kmi_port *kmi = dev_id;
+	unsigned int status = readb(KMIIR);
+	int handled = IRQ_NONE;
+
+	while (status & KMIIR_RXINTR) {
+		serio_interrupt(kmi->io, readb(KMIDATA), 0);
+		status = readb(KMIIR);
+		handled = IRQ_HANDLED;
+	}
+
+	return handled;
+}
+
+static int amba_kmi_write(struct serio *io, unsigned char val)
+{
+	struct amba_kmi_port *kmi = io->port_data;
+	unsigned int timeleft = 10000; /* timeout in 100ms */
+
+	while ((readb(KMISTAT) & KMISTAT_TXEMPTY) == 0 && --timeleft)
+		udelay(10);
+
+	if (timeleft)
+		writeb(val, KMIDATA);
+
+	return timeleft ? 0 : SERIO_TIMEOUT;
+}
+
+static int amba_kmi_open(struct serio *io)
+{
+	struct amba_kmi_port *kmi = io->port_data;
+	unsigned int divisor;
+	int ret;
+
+	ret = clk_prepare_enable(kmi->clk);
+	if (ret)
+		goto out;
+
+	divisor = clk_get_rate(kmi->clk) / 8000000 - 1;
+	writeb(divisor, KMICLKDIV);
+	writeb(KMICR_EN, KMICR);
+
+	ret = request_irq(kmi->irq, amba_kmi_int, IRQF_SHARED, "kmi-pl050",
+			  kmi);
+	if (ret) {
+		printk(KERN_ERR "kmi: failed to claim IRQ%d\n", kmi->irq);
+		writeb(0, KMICR);
+		goto clk_disable;
+	}
+
+	writeb(KMICR_EN | KMICR_RXINTREN, KMICR);
+
+	return 0;
+
+ clk_disable:
+	clk_disable_unprepare(kmi->clk);
+ out:
+	return ret;
+}
+
+static void amba_kmi_close(struct serio *io)
+{
+	struct amba_kmi_port *kmi = io->port_data;
+
+	writeb(0, KMICR);
+
+	free_irq(kmi->irq, kmi);
+	clk_disable_unprepare(kmi->clk);
+}
+
+static int amba_kmi_probe(struct amba_device *dev,
+	const struct amba_id *id)
+{
+	struct amba_kmi_port *kmi;
+	struct serio *io;
+	int ret;
+
+	ret = amba_request_regions(dev, NULL);
+	if (ret)
+		return ret;
+
+	kmi = kzalloc_obj(*kmi);
+	io = kzalloc_obj(*io);
+	if (!kmi || !io) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+
+	io->id.type	= SERIO_8042;
+	io->write	= amba_kmi_write;
+	io->open	= amba_kmi_open;
+	io->close	= amba_kmi_close;
+	strscpy(io->name, dev_name(&dev->dev), sizeof(io->name));
+	strscpy(io->phys, dev_name(&dev->dev), sizeof(io->phys));
+	io->port_data	= kmi;
+	io->dev.parent	= &dev->dev;
+
+	kmi->io		= io;
+	kmi->base	= ioremap(dev->res.start, resource_size(&dev->res));
+	if (!kmi->base) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	kmi->clk = clk_get(&dev->dev, "KMIREFCLK");
+	if (IS_ERR(kmi->clk)) {
+		ret = PTR_ERR(kmi->clk);
+		goto unmap;
+	}
+
+	kmi->irq = dev->irq[0];
+	amba_set_drvdata(dev, kmi);
+
+	serio_register_port(kmi->io);
+	return 0;
+
+ unmap:
+	iounmap(kmi->base);
+ out:
+	kfree(kmi);
+	kfree(io);
+	amba_release_regions(dev);
+	return ret;
+}
+
+static void amba_kmi_remove(struct amba_device *dev)
+{
+	struct amba_kmi_port *kmi = amba_get_drvdata(dev);
+
+	serio_unregister_port(kmi->io);
+	clk_put(kmi->clk);
+	iounmap(kmi->base);
+	kfree(kmi);
+	amba_release_regions(dev);
+}
+
+static int amba_kmi_resume(struct device *dev)
+{
+	struct amba_kmi_port *kmi = dev_get_drvdata(dev);
+
+	/* kick the serio layer to rescan this port */
+	serio_reconnect(kmi->io);
+
+	return 0;
+}
+
+static DEFINE_SIMPLE_DEV_PM_OPS(amba_kmi_dev_pm_ops, NULL, amba_kmi_resume);
+
+static const struct amba_id amba_kmi_idtable[] = {
+	{
+		.id	= 0x00041050,
+		.mask	= 0x000fffff,
+	},
+	{ 0, 0 }
+};
+
+MODULE_DEVICE_TABLE(amba, amba_kmi_idtable);
+
+static struct amba_driver ambakmi_driver = {
+	.drv		= {
+		.name	= "kmi-pl050",
+		.pm	= pm_sleep_ptr(&amba_kmi_dev_pm_ops),
+	},
+	.id_table	= amba_kmi_idtable,
+	.probe		= amba_kmi_probe,
+	.remove		= amba_kmi_remove,
+};
+
+module_amba_driver(ambakmi_driver);
+
+MODULE_AUTHOR("Russell King <rmk@arm.linux.org.uk>");
+MODULE_DESCRIPTION("AMBA KMI controller driver");
+MODULE_LICENSE("GPL");