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, 170 insertions, 0 deletions

diff --git a/drivers/rtc/rtc-ep93xx.c b/drivers/rtc/rtc-ep93xx.c
new file mode 100644
index 000000000..507c9f220
--- /dev/null
+++ b/drivers/rtc/rtc-ep93xx.c
@@ -0,0 +1,170 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * A driver for the RTC embedded in the Cirrus Logic EP93XX processors
+ * Copyright (c) 2006 Tower Technologies
+ *
+ * Author: Alessandro Zummo <a.zummo@towertech.it>
+ */
+
+#include <linux/module.h>
+#include <linux/rtc.h>
+#include <linux/platform_device.h>
+#include <linux/io.h>
+#include <linux/gfp.h>
+
+#define EP93XX_RTC_DATA			0x000
+#define EP93XX_RTC_MATCH		0x004
+#define EP93XX_RTC_STATUS		0x008
+#define  EP93XX_RTC_STATUS_INTR		 BIT(0)
+#define EP93XX_RTC_LOAD			0x00C
+#define EP93XX_RTC_CONTROL		0x010
+#define  EP93XX_RTC_CONTROL_MIE		 BIT(0)
+#define EP93XX_RTC_SWCOMP		0x108
+#define  EP93XX_RTC_SWCOMP_DEL_MASK	 0x001f0000
+#define  EP93XX_RTC_SWCOMP_DEL_SHIFT	 16
+#define  EP93XX_RTC_SWCOMP_INT_MASK	 0x0000ffff
+#define  EP93XX_RTC_SWCOMP_INT_SHIFT	 0
+
+struct ep93xx_rtc {
+	void __iomem	*mmio_base;
+};
+
+static int ep93xx_rtc_get_swcomp(struct device *dev, unsigned short *preload,
+				 unsigned short *delete)
+{
+	struct ep93xx_rtc *ep93xx_rtc = dev_get_drvdata(dev);
+	unsigned long comp;
+
+	comp = readl(ep93xx_rtc->mmio_base + EP93XX_RTC_SWCOMP);
+
+	if (preload)
+		*preload = (comp & EP93XX_RTC_SWCOMP_INT_MASK)
+				>> EP93XX_RTC_SWCOMP_INT_SHIFT;
+
+	if (delete)
+		*delete = (comp & EP93XX_RTC_SWCOMP_DEL_MASK)
+				>> EP93XX_RTC_SWCOMP_DEL_SHIFT;
+
+	return 0;
+}
+
+static int ep93xx_rtc_read_time(struct device *dev, struct rtc_time *tm)
+{
+	struct ep93xx_rtc *ep93xx_rtc = dev_get_drvdata(dev);
+	unsigned long time;
+
+	time = readl(ep93xx_rtc->mmio_base + EP93XX_RTC_DATA);
+
+	rtc_time64_to_tm(time, tm);
+	return 0;
+}
+
+static int ep93xx_rtc_set_time(struct device *dev, struct rtc_time *tm)
+{
+	struct ep93xx_rtc *ep93xx_rtc = dev_get_drvdata(dev);
+	unsigned long secs = rtc_tm_to_time64(tm);
+
+	writel(secs + 1, ep93xx_rtc->mmio_base + EP93XX_RTC_LOAD);
+	return 0;
+}
+
+static int ep93xx_rtc_proc(struct device *dev, struct seq_file *seq)
+{
+	unsigned short preload, delete;
+
+	ep93xx_rtc_get_swcomp(dev, &preload, &delete);
+
+	seq_printf(seq, "preload\t\t: %d\n", preload);
+	seq_printf(seq, "delete\t\t: %d\n", delete);
+
+	return 0;
+}
+
+static const struct rtc_class_ops ep93xx_rtc_ops = {
+	.read_time	= ep93xx_rtc_read_time,
+	.set_time	= ep93xx_rtc_set_time,
+	.proc		= ep93xx_rtc_proc,
+};
+
+static ssize_t comp_preload_show(struct device *dev,
+				 struct device_attribute *attr, char *buf)
+{
+	unsigned short preload;
+
+	ep93xx_rtc_get_swcomp(dev->parent, &preload, NULL);
+
+	return sprintf(buf, "%d\n", preload);
+}
+static DEVICE_ATTR_RO(comp_preload);
+
+static ssize_t comp_delete_show(struct device *dev,
+				struct device_attribute *attr, char *buf)
+{
+	unsigned short delete;
+
+	ep93xx_rtc_get_swcomp(dev->parent, NULL, &delete);
+
+	return sprintf(buf, "%d\n", delete);
+}
+static DEVICE_ATTR_RO(comp_delete);
+
+static struct attribute *ep93xx_rtc_attrs[] = {
+	&dev_attr_comp_preload.attr,
+	&dev_attr_comp_delete.attr,
+	NULL
+};
+
+static const struct attribute_group ep93xx_rtc_sysfs_files = {
+	.attrs	= ep93xx_rtc_attrs,
+};
+
+static int ep93xx_rtc_probe(struct platform_device *pdev)
+{
+	struct ep93xx_rtc *ep93xx_rtc;
+	struct rtc_device *rtc;
+	int err;
+
+	ep93xx_rtc = devm_kzalloc(&pdev->dev, sizeof(*ep93xx_rtc), GFP_KERNEL);
+	if (!ep93xx_rtc)
+		return -ENOMEM;
+
+	ep93xx_rtc->mmio_base = devm_platform_ioremap_resource(pdev, 0);
+	if (IS_ERR(ep93xx_rtc->mmio_base))
+		return PTR_ERR(ep93xx_rtc->mmio_base);
+
+	platform_set_drvdata(pdev, ep93xx_rtc);
+
+	rtc = devm_rtc_allocate_device(&pdev->dev);
+	if (IS_ERR(rtc))
+		return PTR_ERR(rtc);
+
+	rtc->ops = &ep93xx_rtc_ops;
+	rtc->range_max = U32_MAX;
+
+	err = rtc_add_group(rtc, &ep93xx_rtc_sysfs_files);
+	if (err)
+		return err;
+
+	return devm_rtc_register_device(rtc);
+}
+
+static const struct of_device_id ep93xx_rtc_of_ids[] = {
+	{ .compatible = "cirrus,ep9301-rtc" },
+	{ /* sentinel */ }
+};
+MODULE_DEVICE_TABLE(of, ep93xx_rtc_of_ids);
+
+static struct platform_driver ep93xx_rtc_driver = {
+	.driver		= {
+		.name	= "ep93xx-rtc",
+		.of_match_table = ep93xx_rtc_of_ids,
+	},
+	.probe		= ep93xx_rtc_probe,
+};
+
+module_platform_driver(ep93xx_rtc_driver);
+
+MODULE_AUTHOR("Alessandro Zummo <a.zummo@towertech.it>");
+MODULE_DESCRIPTION("EP93XX RTC driver");
+MODULE_LICENSE("GPL");
+MODULE_ALIAS("platform:ep93xx-rtc");