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, 322 insertions, 0 deletions

diff --git a/drivers/rtc/rtc-lp8788.c b/drivers/rtc/rtc-lp8788.c
new file mode 100644
index 000000000..0793d7050
--- /dev/null
+++ b/drivers/rtc/rtc-lp8788.c
@@ -0,0 +1,322 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * TI LP8788 MFD - rtc driver
+ *
+ * Copyright 2012 Texas Instruments
+ *
+ * Author: Milo(Woogyom) Kim <milo.kim@ti.com>
+ */
+
+#include <linux/err.h>
+#include <linux/irqdomain.h>
+#include <linux/mfd/lp8788.h>
+#include <linux/module.h>
+#include <linux/platform_device.h>
+#include <linux/rtc.h>
+#include <linux/slab.h>
+
+/* register address */
+#define LP8788_INTEN_3			0x05
+#define LP8788_RTC_UNLOCK		0x64
+#define LP8788_RTC_SEC			0x70
+#define LP8788_ALM1_SEC			0x77
+#define LP8788_ALM1_EN			0x7D
+#define LP8788_ALM2_SEC			0x7E
+#define LP8788_ALM2_EN			0x84
+
+/* mask/shift bits */
+#define LP8788_INT_RTC_ALM1_M		BIT(1)	/* Addr 05h */
+#define LP8788_INT_RTC_ALM1_S		1
+#define LP8788_INT_RTC_ALM2_M		BIT(2)	/* Addr 05h */
+#define LP8788_INT_RTC_ALM2_S		2
+#define LP8788_ALM_EN_M			BIT(7)	/* Addr 7Dh or 84h */
+#define LP8788_ALM_EN_S			7
+
+#define DEFAULT_ALARM_SEL		LP8788_ALARM_1
+#define LP8788_MONTH_OFFSET		1
+#define LP8788_BASE_YEAR		2000
+#define MAX_WDAY_BITS			7
+#define LP8788_WDAY_SET			1
+#define RTC_UNLOCK			0x1
+#define RTC_LATCH			0x2
+#define ALARM_IRQ_FLAG			(RTC_IRQF | RTC_AF)
+
+enum lp8788_time {
+	LPTIME_SEC,
+	LPTIME_MIN,
+	LPTIME_HOUR,
+	LPTIME_MDAY,
+	LPTIME_MON,
+	LPTIME_YEAR,
+	LPTIME_WDAY,
+	LPTIME_MAX,
+};
+
+struct lp8788_rtc {
+	struct lp8788 *lp;
+	struct rtc_device *rdev;
+	enum lp8788_alarm_sel alarm;
+	int irq;
+};
+
+static const u8 addr_alarm_sec[LP8788_ALARM_MAX] = {
+	LP8788_ALM1_SEC,
+	LP8788_ALM2_SEC,
+};
+
+static const u8 addr_alarm_en[LP8788_ALARM_MAX] = {
+	LP8788_ALM1_EN,
+	LP8788_ALM2_EN,
+};
+
+static const u8 mask_alarm_en[LP8788_ALARM_MAX] = {
+	LP8788_INT_RTC_ALM1_M,
+	LP8788_INT_RTC_ALM2_M,
+};
+
+static const u8 shift_alarm_en[LP8788_ALARM_MAX] = {
+	LP8788_INT_RTC_ALM1_S,
+	LP8788_INT_RTC_ALM2_S,
+};
+
+static int _to_tm_wday(u8 lp8788_wday)
+{
+	int i;
+
+	if (lp8788_wday == 0)
+		return 0;
+
+	/* lookup defined weekday from read register value */
+	for (i = 0; i < MAX_WDAY_BITS; i++) {
+		if ((lp8788_wday >> i) == LP8788_WDAY_SET)
+			break;
+	}
+
+	return i + 1;
+}
+
+static inline int _to_lp8788_wday(int tm_wday)
+{
+	return LP8788_WDAY_SET << (tm_wday - 1);
+}
+
+static void lp8788_rtc_unlock(struct lp8788 *lp)
+{
+	lp8788_write_byte(lp, LP8788_RTC_UNLOCK, RTC_UNLOCK);
+	lp8788_write_byte(lp, LP8788_RTC_UNLOCK, RTC_LATCH);
+}
+
+static int lp8788_rtc_read_time(struct device *dev, struct rtc_time *tm)
+{
+	struct lp8788_rtc *rtc = dev_get_drvdata(dev);
+	struct lp8788 *lp = rtc->lp;
+	u8 data[LPTIME_MAX];
+	int ret;
+
+	lp8788_rtc_unlock(lp);
+
+	ret = lp8788_read_multi_bytes(lp, LP8788_RTC_SEC, data,	LPTIME_MAX);
+	if (ret)
+		return ret;
+
+	tm->tm_sec  = data[LPTIME_SEC];
+	tm->tm_min  = data[LPTIME_MIN];
+	tm->tm_hour = data[LPTIME_HOUR];
+	tm->tm_mday = data[LPTIME_MDAY];
+	tm->tm_mon  = data[LPTIME_MON] - LP8788_MONTH_OFFSET;
+	tm->tm_year = data[LPTIME_YEAR] + LP8788_BASE_YEAR - 1900;
+	tm->tm_wday = _to_tm_wday(data[LPTIME_WDAY]);
+
+	return 0;
+}
+
+static int lp8788_rtc_set_time(struct device *dev, struct rtc_time *tm)
+{
+	struct lp8788_rtc *rtc = dev_get_drvdata(dev);
+	struct lp8788 *lp = rtc->lp;
+	u8 data[LPTIME_MAX - 1];
+	int ret, i, year;
+
+	year = tm->tm_year + 1900 - LP8788_BASE_YEAR;
+	if (year < 0) {
+		dev_err(lp->dev, "invalid year: %d\n", year);
+		return -EINVAL;
+	}
+
+	/* because rtc weekday is a readonly register, do not update */
+	data[LPTIME_SEC]  = tm->tm_sec;
+	data[LPTIME_MIN]  = tm->tm_min;
+	data[LPTIME_HOUR] = tm->tm_hour;
+	data[LPTIME_MDAY] = tm->tm_mday;
+	data[LPTIME_MON]  = tm->tm_mon + LP8788_MONTH_OFFSET;
+	data[LPTIME_YEAR] = year;
+
+	for (i = 0; i < ARRAY_SIZE(data); i++) {
+		ret = lp8788_write_byte(lp, LP8788_RTC_SEC + i, data[i]);
+		if (ret)
+			return ret;
+	}
+
+	return 0;
+}
+
+static int lp8788_read_alarm(struct device *dev, struct rtc_wkalrm *alarm)
+{
+	struct lp8788_rtc *rtc = dev_get_drvdata(dev);
+	struct lp8788 *lp = rtc->lp;
+	struct rtc_time *tm = &alarm->time;
+	u8 addr, data[LPTIME_MAX];
+	int ret;
+
+	addr = addr_alarm_sec[rtc->alarm];
+	ret = lp8788_read_multi_bytes(lp, addr, data, LPTIME_MAX);
+	if (ret)
+		return ret;
+
+	tm->tm_sec  = data[LPTIME_SEC];
+	tm->tm_min  = data[LPTIME_MIN];
+	tm->tm_hour = data[LPTIME_HOUR];
+	tm->tm_mday = data[LPTIME_MDAY];
+	tm->tm_mon  = data[LPTIME_MON] - LP8788_MONTH_OFFSET;
+	tm->tm_year = data[LPTIME_YEAR] + LP8788_BASE_YEAR - 1900;
+	tm->tm_wday = _to_tm_wday(data[LPTIME_WDAY]);
+	alarm->enabled = data[LPTIME_WDAY] & LP8788_ALM_EN_M;
+
+	return 0;
+}
+
+static int lp8788_set_alarm(struct device *dev, struct rtc_wkalrm *alarm)
+{
+	struct lp8788_rtc *rtc = dev_get_drvdata(dev);
+	struct lp8788 *lp = rtc->lp;
+	struct rtc_time *tm = &alarm->time;
+	u8 addr, data[LPTIME_MAX];
+	int ret, i, year;
+
+	year = tm->tm_year + 1900 - LP8788_BASE_YEAR;
+	if (year < 0) {
+		dev_err(lp->dev, "invalid year: %d\n", year);
+		return -EINVAL;
+	}
+
+	data[LPTIME_SEC]  = tm->tm_sec;
+	data[LPTIME_MIN]  = tm->tm_min;
+	data[LPTIME_HOUR] = tm->tm_hour;
+	data[LPTIME_MDAY] = tm->tm_mday;
+	data[LPTIME_MON]  = tm->tm_mon + LP8788_MONTH_OFFSET;
+	data[LPTIME_YEAR] = year;
+	data[LPTIME_WDAY] = _to_lp8788_wday(tm->tm_wday);
+
+	for (i = 0; i < ARRAY_SIZE(data); i++) {
+		addr = addr_alarm_sec[rtc->alarm] + i;
+		ret = lp8788_write_byte(lp, addr, data[i]);
+		if (ret)
+			return ret;
+	}
+
+	alarm->enabled = 1;
+	addr = addr_alarm_en[rtc->alarm];
+
+	return lp8788_update_bits(lp, addr, LP8788_ALM_EN_M,
+				alarm->enabled << LP8788_ALM_EN_S);
+}
+
+static int lp8788_alarm_irq_enable(struct device *dev, unsigned int enable)
+{
+	struct lp8788_rtc *rtc = dev_get_drvdata(dev);
+	struct lp8788 *lp = rtc->lp;
+	u8 mask, shift;
+
+	if (!rtc->irq)
+		return -EIO;
+
+	mask = mask_alarm_en[rtc->alarm];
+	shift = shift_alarm_en[rtc->alarm];
+
+	return lp8788_update_bits(lp, LP8788_INTEN_3, mask, enable << shift);
+}
+
+static const struct rtc_class_ops lp8788_rtc_ops = {
+	.read_time = lp8788_rtc_read_time,
+	.set_time = lp8788_rtc_set_time,
+	.read_alarm = lp8788_read_alarm,
+	.set_alarm = lp8788_set_alarm,
+	.alarm_irq_enable = lp8788_alarm_irq_enable,
+};
+
+static irqreturn_t lp8788_alarm_irq_handler(int irq, void *ptr)
+{
+	struct lp8788_rtc *rtc = ptr;
+
+	rtc_update_irq(rtc->rdev, 1, ALARM_IRQ_FLAG);
+	return IRQ_HANDLED;
+}
+
+static int lp8788_alarm_irq_register(struct platform_device *pdev,
+				struct lp8788_rtc *rtc)
+{
+	struct resource *r;
+	struct lp8788 *lp = rtc->lp;
+	struct irq_domain *irqdm = lp->irqdm;
+	int irq;
+
+	rtc->irq = 0;
+
+	/* even the alarm IRQ number is not specified, rtc time should work */
+	r = platform_get_resource_byname(pdev, IORESOURCE_IRQ, LP8788_ALM_IRQ);
+	if (!r)
+		return 0;
+
+	if (rtc->alarm == LP8788_ALARM_1)
+		irq = r->start;
+	else
+		irq = r->end;
+
+	rtc->irq = irq_create_mapping(irqdm, irq);
+
+	return devm_request_threaded_irq(&pdev->dev, rtc->irq, NULL,
+				lp8788_alarm_irq_handler,
+				0, LP8788_ALM_IRQ, rtc);
+}
+
+static int lp8788_rtc_probe(struct platform_device *pdev)
+{
+	struct lp8788 *lp = dev_get_drvdata(pdev->dev.parent);
+	struct lp8788_rtc *rtc;
+	struct device *dev = &pdev->dev;
+
+	rtc = devm_kzalloc(dev, sizeof(struct lp8788_rtc), GFP_KERNEL);
+	if (!rtc)
+		return -ENOMEM;
+
+	rtc->lp = lp;
+	rtc->alarm = lp->pdata ? lp->pdata->alarm_sel : DEFAULT_ALARM_SEL;
+	platform_set_drvdata(pdev, rtc);
+
+	device_init_wakeup(dev, true);
+
+	rtc->rdev = devm_rtc_device_register(dev, "lp8788_rtc",
+					&lp8788_rtc_ops, THIS_MODULE);
+	if (IS_ERR(rtc->rdev)) {
+		dev_err(dev, "can not register rtc device\n");
+		return PTR_ERR(rtc->rdev);
+	}
+
+	if (lp8788_alarm_irq_register(pdev, rtc))
+		dev_warn(lp->dev, "no rtc irq handler\n");
+
+	return 0;
+}
+
+static struct platform_driver lp8788_rtc_driver = {
+	.probe = lp8788_rtc_probe,
+	.driver = {
+		.name = LP8788_DEV_RTC,
+	},
+};
+module_platform_driver(lp8788_rtc_driver);
+
+MODULE_DESCRIPTION("Texas Instruments LP8788 RTC Driver");
+MODULE_AUTHOR("Milo Kim");
+MODULE_LICENSE("GPL");
+MODULE_ALIAS("platform:lp8788-rtc");