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, 307 insertions, 0 deletions

diff --git a/drivers/rtc/rtc-moxart.c b/drivers/rtc/rtc-moxart.c
new file mode 100644
index 000000000..bc0345db6
--- /dev/null
+++ b/drivers/rtc/rtc-moxart.c
@@ -0,0 +1,307 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * MOXA ART RTC driver.
+ *
+ * Copyright (C) 2013 Jonas Jensen
+ *
+ * Jonas Jensen <jonas.jensen@gmail.com>
+ *
+ * Based on code from
+ * Moxa Technology Co., Ltd. <www.moxa.com>
+ */
+
+#include <linux/err.h>
+#include <linux/init.h>
+#include <linux/kernel.h>
+#include <linux/delay.h>
+#include <linux/rtc.h>
+#include <linux/platform_device.h>
+#include <linux/module.h>
+#include <linux/gpio/consumer.h>
+
+#define GPIO_RTC_RESERVED			0x0C
+#define GPIO_RTC_DATA_SET			0x10
+#define GPIO_RTC_DATA_CLEAR			0x14
+#define GPIO_RTC_PIN_PULL_ENABLE		0x18
+#define GPIO_RTC_PIN_PULL_TYPE			0x1C
+#define GPIO_RTC_INT_ENABLE			0x20
+#define GPIO_RTC_INT_RAW_STATE			0x24
+#define GPIO_RTC_INT_MASKED_STATE		0x28
+#define GPIO_RTC_INT_MASK			0x2C
+#define GPIO_RTC_INT_CLEAR			0x30
+#define GPIO_RTC_INT_TRIGGER			0x34
+#define GPIO_RTC_INT_BOTH			0x38
+#define GPIO_RTC_INT_RISE_NEG			0x3C
+#define GPIO_RTC_BOUNCE_ENABLE			0x40
+#define GPIO_RTC_BOUNCE_PRE_SCALE		0x44
+#define GPIO_RTC_PROTECT_W			0x8E
+#define GPIO_RTC_PROTECT_R			0x8F
+#define GPIO_RTC_YEAR_W				0x8C
+#define GPIO_RTC_YEAR_R				0x8D
+#define GPIO_RTC_DAY_W				0x8A
+#define GPIO_RTC_DAY_R				0x8B
+#define GPIO_RTC_MONTH_W			0x88
+#define GPIO_RTC_MONTH_R			0x89
+#define GPIO_RTC_DATE_W				0x86
+#define GPIO_RTC_DATE_R				0x87
+#define GPIO_RTC_HOURS_W			0x84
+#define GPIO_RTC_HOURS_R			0x85
+#define GPIO_RTC_MINUTES_W			0x82
+#define GPIO_RTC_MINUTES_R			0x83
+#define GPIO_RTC_SECONDS_W			0x80
+#define GPIO_RTC_SECONDS_R			0x81
+#define GPIO_RTC_DELAY_TIME			8
+
+struct moxart_rtc {
+	struct rtc_device *rtc;
+	spinlock_t rtc_lock;
+	struct gpio_desc *gpio_data;
+	struct gpio_desc *gpio_sclk;
+	struct gpio_desc *gpio_reset;
+};
+
+static int day_of_year[12] =	{ 0, 31, 59, 90, 120, 151, 181,
+				  212, 243, 273, 304, 334 };
+
+static void moxart_rtc_write_byte(struct device *dev, u8 data)
+{
+	struct moxart_rtc *moxart_rtc = dev_get_drvdata(dev);
+	int i;
+
+	for (i = 0; i < 8; i++, data >>= 1) {
+		gpiod_set_value(moxart_rtc->gpio_sclk, 0);
+		gpiod_set_value(moxart_rtc->gpio_data, ((data & 1) == 1));
+		udelay(GPIO_RTC_DELAY_TIME);
+		gpiod_set_value(moxart_rtc->gpio_sclk, 1);
+		udelay(GPIO_RTC_DELAY_TIME);
+	}
+}
+
+static u8 moxart_rtc_read_byte(struct device *dev)
+{
+	struct moxart_rtc *moxart_rtc = dev_get_drvdata(dev);
+	int i;
+	u8 data = 0;
+
+	for (i = 0; i < 8; i++) {
+		gpiod_set_value(moxart_rtc->gpio_sclk, 0);
+		udelay(GPIO_RTC_DELAY_TIME);
+		gpiod_set_value(moxart_rtc->gpio_sclk, 1);
+		udelay(GPIO_RTC_DELAY_TIME);
+		if (gpiod_get_value(moxart_rtc->gpio_data))
+			data |= (1 << i);
+		udelay(GPIO_RTC_DELAY_TIME);
+	}
+	return data;
+}
+
+static u8 moxart_rtc_read_register(struct device *dev, u8 cmd)
+{
+	struct moxart_rtc *moxart_rtc = dev_get_drvdata(dev);
+	u8 data;
+	unsigned long flags;
+
+	local_irq_save(flags);
+
+	gpiod_direction_output(moxart_rtc->gpio_data, 0);
+	gpiod_set_value(moxart_rtc->gpio_reset, 1);
+	udelay(GPIO_RTC_DELAY_TIME);
+	moxart_rtc_write_byte(dev, cmd);
+	gpiod_direction_input(moxart_rtc->gpio_data);
+	udelay(GPIO_RTC_DELAY_TIME);
+	data = moxart_rtc_read_byte(dev);
+	gpiod_set_value(moxart_rtc->gpio_sclk, 0);
+	gpiod_set_value(moxart_rtc->gpio_reset, 0);
+	udelay(GPIO_RTC_DELAY_TIME);
+
+	local_irq_restore(flags);
+
+	return data;
+}
+
+static void moxart_rtc_write_register(struct device *dev, u8 cmd, u8 data)
+{
+	struct moxart_rtc *moxart_rtc = dev_get_drvdata(dev);
+	unsigned long flags;
+
+	local_irq_save(flags);
+
+	gpiod_direction_output(moxart_rtc->gpio_data, 0);
+	gpiod_set_value(moxart_rtc->gpio_reset, 1);
+	udelay(GPIO_RTC_DELAY_TIME);
+	moxart_rtc_write_byte(dev, cmd);
+	moxart_rtc_write_byte(dev, data);
+	gpiod_set_value(moxart_rtc->gpio_sclk, 0);
+	gpiod_set_value(moxart_rtc->gpio_reset, 0);
+	udelay(GPIO_RTC_DELAY_TIME);
+
+	local_irq_restore(flags);
+}
+
+static int moxart_rtc_set_time(struct device *dev, struct rtc_time *tm)
+{
+	struct moxart_rtc *moxart_rtc = dev_get_drvdata(dev);
+
+	spin_lock_irq(&moxart_rtc->rtc_lock);
+
+	moxart_rtc_write_register(dev, GPIO_RTC_PROTECT_W, 0);
+	moxart_rtc_write_register(dev, GPIO_RTC_YEAR_W,
+				  (((tm->tm_year - 100) / 10) << 4) |
+				  ((tm->tm_year - 100) % 10));
+
+	moxart_rtc_write_register(dev, GPIO_RTC_MONTH_W,
+				  (((tm->tm_mon + 1) / 10) << 4) |
+				  ((tm->tm_mon + 1) % 10));
+
+	moxart_rtc_write_register(dev, GPIO_RTC_DATE_W,
+				  ((tm->tm_mday / 10) << 4) |
+				  (tm->tm_mday % 10));
+
+	moxart_rtc_write_register(dev, GPIO_RTC_HOURS_W,
+				  ((tm->tm_hour / 10) << 4) |
+				  (tm->tm_hour % 10));
+
+	moxart_rtc_write_register(dev, GPIO_RTC_MINUTES_W,
+				  ((tm->tm_min / 10) << 4) |
+				  (tm->tm_min % 10));
+
+	moxart_rtc_write_register(dev, GPIO_RTC_SECONDS_W,
+				  ((tm->tm_sec / 10) << 4) |
+				  (tm->tm_sec % 10));
+
+	moxart_rtc_write_register(dev, GPIO_RTC_PROTECT_W, 0x80);
+
+	spin_unlock_irq(&moxart_rtc->rtc_lock);
+
+	dev_dbg(dev, "%s: success tm_year=%d tm_mon=%d\n"
+		"tm_mday=%d tm_hour=%d tm_min=%d tm_sec=%d\n",
+		__func__, tm->tm_year, tm->tm_mon, tm->tm_mday,
+		tm->tm_hour, tm->tm_min, tm->tm_sec);
+
+	return 0;
+}
+
+static int moxart_rtc_read_time(struct device *dev, struct rtc_time *tm)
+{
+	struct moxart_rtc *moxart_rtc = dev_get_drvdata(dev);
+	unsigned char v;
+
+	spin_lock_irq(&moxart_rtc->rtc_lock);
+
+	v = moxart_rtc_read_register(dev, GPIO_RTC_SECONDS_R);
+	tm->tm_sec = (((v & 0x70) >> 4) * 10) + (v & 0x0F);
+
+	v = moxart_rtc_read_register(dev, GPIO_RTC_MINUTES_R);
+	tm->tm_min = (((v & 0x70) >> 4) * 10) + (v & 0x0F);
+
+	v = moxart_rtc_read_register(dev, GPIO_RTC_HOURS_R);
+	if (v & 0x80) { /* 12-hour mode */
+		tm->tm_hour = (((v & 0x10) >> 4) * 10) + (v & 0x0F);
+		if (v & 0x20) { /* PM mode */
+			tm->tm_hour += 12;
+			if (tm->tm_hour >= 24)
+				tm->tm_hour = 0;
+		}
+	} else { /* 24-hour mode */
+		tm->tm_hour = (((v & 0x30) >> 4) * 10) + (v & 0x0F);
+	}
+
+	v = moxart_rtc_read_register(dev, GPIO_RTC_DATE_R);
+	tm->tm_mday = (((v & 0x30) >> 4) * 10) + (v & 0x0F);
+
+	v = moxart_rtc_read_register(dev, GPIO_RTC_MONTH_R);
+	tm->tm_mon = (((v & 0x10) >> 4) * 10) + (v & 0x0F);
+	tm->tm_mon--;
+
+	v = moxart_rtc_read_register(dev, GPIO_RTC_YEAR_R);
+	tm->tm_year = (((v & 0xF0) >> 4) * 10) + (v & 0x0F);
+	tm->tm_year += 100;
+	if (tm->tm_year <= 69)
+		tm->tm_year += 100;
+
+	v = moxart_rtc_read_register(dev, GPIO_RTC_DAY_R);
+	tm->tm_wday = (v & 0x0f) - 1;
+	tm->tm_yday = day_of_year[tm->tm_mon];
+	tm->tm_yday += (tm->tm_mday - 1);
+	if (tm->tm_mon >= 2) {
+		if (!(tm->tm_year % 4) && (tm->tm_year % 100))
+			tm->tm_yday++;
+	}
+
+	tm->tm_isdst = 0;
+
+	spin_unlock_irq(&moxart_rtc->rtc_lock);
+
+	return 0;
+}
+
+static const struct rtc_class_ops moxart_rtc_ops = {
+	.read_time	= moxart_rtc_read_time,
+	.set_time	= moxart_rtc_set_time,
+};
+
+static int moxart_rtc_probe(struct platform_device *pdev)
+{
+	struct moxart_rtc *moxart_rtc;
+	int ret = 0;
+
+	moxart_rtc = devm_kzalloc(&pdev->dev, sizeof(*moxart_rtc), GFP_KERNEL);
+	if (!moxart_rtc)
+		return -ENOMEM;
+
+	moxart_rtc->gpio_data = devm_gpiod_get(&pdev->dev, "rtc-data",
+					       GPIOD_IN);
+	ret = PTR_ERR_OR_ZERO(moxart_rtc->gpio_data);
+	if (ret) {
+		dev_err(&pdev->dev, "can't get rtc data gpio: %d\n", ret);
+		return ret;
+	}
+
+	moxart_rtc->gpio_sclk = devm_gpiod_get(&pdev->dev, "rtc-sclk",
+					       GPIOD_ASIS);
+	ret = PTR_ERR_OR_ZERO(moxart_rtc->gpio_sclk);
+	if (ret) {
+		dev_err(&pdev->dev, "can't get rtc sclk gpio: %d\n", ret);
+		return ret;
+	}
+
+	moxart_rtc->gpio_reset = devm_gpiod_get(&pdev->dev, "rtc-reset",
+						GPIOD_ASIS);
+	ret = PTR_ERR_OR_ZERO(moxart_rtc->gpio_reset);
+	if (ret) {
+		dev_err(&pdev->dev, "can't get rtc reset gpio: %d\n", ret);
+		return ret;
+	}
+
+	spin_lock_init(&moxart_rtc->rtc_lock);
+	platform_set_drvdata(pdev, moxart_rtc);
+
+	moxart_rtc->rtc = devm_rtc_device_register(&pdev->dev, pdev->name,
+						   &moxart_rtc_ops,
+						   THIS_MODULE);
+	if (IS_ERR(moxart_rtc->rtc)) {
+		dev_err(&pdev->dev, "devm_rtc_device_register failed\n");
+		return PTR_ERR(moxart_rtc->rtc);
+	}
+
+	return 0;
+}
+
+static const struct of_device_id moxart_rtc_match[] = {
+	{ .compatible = "moxa,moxart-rtc" },
+	{ },
+};
+MODULE_DEVICE_TABLE(of, moxart_rtc_match);
+
+static struct platform_driver moxart_rtc_driver = {
+	.probe	= moxart_rtc_probe,
+	.driver	= {
+		.name		= "moxart-rtc",
+		.of_match_table	= moxart_rtc_match,
+	},
+};
+module_platform_driver(moxart_rtc_driver);
+
+MODULE_DESCRIPTION("MOXART RTC driver");
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Jonas Jensen <jonas.jensen@gmail.com>");