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, 218 insertions, 0 deletions

diff --git a/drivers/rtc/rtc-ftrtc010.c b/drivers/rtc/rtc-ftrtc010.c
new file mode 100644
index 000000000..98fa0f518
--- /dev/null
+++ b/drivers/rtc/rtc-ftrtc010.c
@@ -0,0 +1,218 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ *  Faraday Technology FTRTC010 driver
+ *
+ *  Copyright (C) 2009 Janos Laube <janos.dev@gmail.com>
+ *
+ * Original code for older kernel 2.6.15 are from Stormlinksemi
+ * first update from Janos Laube for > 2.6.29 kernels
+ *
+ * checkpatch fixes and usage of rtc-lib code
+ * Hans Ulli Kroll <ulli.kroll@googlemail.com>
+ */
+
+#include <linux/rtc.h>
+#include <linux/io.h>
+#include <linux/slab.h>
+#include <linux/platform_device.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/clk.h>
+
+#define DRV_NAME        "rtc-ftrtc010"
+
+MODULE_AUTHOR("Hans Ulli Kroll <ulli.kroll@googlemail.com>");
+MODULE_DESCRIPTION("RTC driver for Gemini SoC");
+MODULE_LICENSE("GPL");
+MODULE_ALIAS("platform:" DRV_NAME);
+
+struct ftrtc010_rtc {
+	void __iomem		*rtc_base;
+	int			rtc_irq;
+	struct clk		*pclk;
+	struct clk		*extclk;
+};
+
+enum ftrtc010_rtc_offsets {
+	FTRTC010_RTC_SECOND		= 0x00,
+	FTRTC010_RTC_MINUTE		= 0x04,
+	FTRTC010_RTC_HOUR		= 0x08,
+	FTRTC010_RTC_DAYS		= 0x0C,
+	FTRTC010_RTC_ALARM_SECOND	= 0x10,
+	FTRTC010_RTC_ALARM_MINUTE	= 0x14,
+	FTRTC010_RTC_ALARM_HOUR		= 0x18,
+	FTRTC010_RTC_RECORD		= 0x1C,
+	FTRTC010_RTC_CR			= 0x20,
+};
+
+static irqreturn_t ftrtc010_rtc_interrupt(int irq, void *dev)
+{
+	return IRQ_HANDLED;
+}
+
+/*
+ * Looks like the RTC in the Gemini SoC is (totaly) broken
+ * We can't read/write directly the time from RTC registers.
+ * We must do some "offset" calculation to get the real time
+ *
+ * This FIX works pretty fine and Stormlinksemi aka Cortina-Networks does
+ * the same thing, without the rtc-lib.c calls.
+ */
+
+static int ftrtc010_rtc_read_time(struct device *dev, struct rtc_time *tm)
+{
+	struct ftrtc010_rtc *rtc = dev_get_drvdata(dev);
+
+	u32 days, hour, min, sec, offset;
+	timeu64_t time;
+
+	sec  = readl(rtc->rtc_base + FTRTC010_RTC_SECOND);
+	min  = readl(rtc->rtc_base + FTRTC010_RTC_MINUTE);
+	hour = readl(rtc->rtc_base + FTRTC010_RTC_HOUR);
+	days = readl(rtc->rtc_base + FTRTC010_RTC_DAYS);
+	offset = readl(rtc->rtc_base + FTRTC010_RTC_RECORD);
+
+	time = offset + days * 86400 + hour * 3600 + min * 60 + sec;
+
+	rtc_time64_to_tm(time, tm);
+
+	return 0;
+}
+
+static int ftrtc010_rtc_set_time(struct device *dev, struct rtc_time *tm)
+{
+	struct ftrtc010_rtc *rtc = dev_get_drvdata(dev);
+	u32 sec, min, hour, day, offset;
+	timeu64_t time;
+
+	time = rtc_tm_to_time64(tm);
+
+	sec = readl(rtc->rtc_base + FTRTC010_RTC_SECOND);
+	min = readl(rtc->rtc_base + FTRTC010_RTC_MINUTE);
+	hour = readl(rtc->rtc_base + FTRTC010_RTC_HOUR);
+	day = readl(rtc->rtc_base + FTRTC010_RTC_DAYS);
+
+	offset = time - (day * 86400 + hour * 3600 + min * 60 + sec);
+
+	writel(offset, rtc->rtc_base + FTRTC010_RTC_RECORD);
+	writel(0x01, rtc->rtc_base + FTRTC010_RTC_CR);
+
+	return 0;
+}
+
+static const struct rtc_class_ops ftrtc010_rtc_ops = {
+	.read_time     = ftrtc010_rtc_read_time,
+	.set_time      = ftrtc010_rtc_set_time,
+};
+
+static int ftrtc010_rtc_probe(struct platform_device *pdev)
+{
+	u32 days, hour, min, sec;
+	struct ftrtc010_rtc *rtc;
+	struct device *dev = &pdev->dev;
+	struct resource *res;
+	struct rtc_device *rtc_dev;
+	int ret;
+
+	rtc = devm_kzalloc(&pdev->dev, sizeof(*rtc), GFP_KERNEL);
+	if (unlikely(!rtc))
+		return -ENOMEM;
+	platform_set_drvdata(pdev, rtc);
+
+	rtc->pclk = devm_clk_get(dev, "PCLK");
+	if (IS_ERR(rtc->pclk)) {
+		dev_err(dev, "could not get PCLK\n");
+	} else {
+		ret = clk_prepare_enable(rtc->pclk);
+		if (ret) {
+			dev_err(dev, "failed to enable PCLK\n");
+			return ret;
+		}
+	}
+	rtc->extclk = devm_clk_get(dev, "EXTCLK");
+	if (IS_ERR(rtc->extclk)) {
+		dev_err(dev, "could not get EXTCLK\n");
+	} else {
+		ret = clk_prepare_enable(rtc->extclk);
+		if (ret) {
+			dev_err(dev, "failed to enable EXTCLK\n");
+			goto err_disable_pclk;
+		}
+	}
+
+	rtc->rtc_irq = platform_get_irq(pdev, 0);
+	if (rtc->rtc_irq < 0) {
+		ret = rtc->rtc_irq;
+		goto err_disable_extclk;
+	}
+
+	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+	if (!res) {
+		ret = -ENODEV;
+		goto err_disable_extclk;
+	}
+
+	rtc->rtc_base = devm_ioremap(dev, res->start,
+				     resource_size(res));
+	if (!rtc->rtc_base) {
+		ret = -ENOMEM;
+		goto err_disable_extclk;
+	}
+
+	rtc_dev = devm_rtc_allocate_device(dev);
+	if (IS_ERR(rtc_dev)) {
+		ret = PTR_ERR(rtc_dev);
+		goto err_disable_extclk;
+	}
+
+	rtc_dev->ops = &ftrtc010_rtc_ops;
+
+	sec  = readl(rtc->rtc_base + FTRTC010_RTC_SECOND);
+	min  = readl(rtc->rtc_base + FTRTC010_RTC_MINUTE);
+	hour = readl(rtc->rtc_base + FTRTC010_RTC_HOUR);
+	days = readl(rtc->rtc_base + FTRTC010_RTC_DAYS);
+
+	rtc_dev->range_min = (u64)days * 86400 + hour * 3600 + min * 60 + sec;
+	rtc_dev->range_max = U32_MAX + rtc_dev->range_min;
+
+	ret = devm_request_irq(dev, rtc->rtc_irq, ftrtc010_rtc_interrupt,
+			       IRQF_SHARED, pdev->name, dev);
+	if (unlikely(ret))
+		goto err_disable_extclk;
+
+	return devm_rtc_register_device(rtc_dev);
+
+err_disable_extclk:
+	clk_disable_unprepare(rtc->extclk);
+err_disable_pclk:
+	clk_disable_unprepare(rtc->pclk);
+	return ret;
+}
+
+static void ftrtc010_rtc_remove(struct platform_device *pdev)
+{
+	struct ftrtc010_rtc *rtc = platform_get_drvdata(pdev);
+
+	if (!IS_ERR(rtc->extclk))
+		clk_disable_unprepare(rtc->extclk);
+	if (!IS_ERR(rtc->pclk))
+		clk_disable_unprepare(rtc->pclk);
+}
+
+static const struct of_device_id ftrtc010_rtc_dt_match[] = {
+	{ .compatible = "cortina,gemini-rtc" },
+	{ .compatible = "faraday,ftrtc010" },
+	{ }
+};
+MODULE_DEVICE_TABLE(of, ftrtc010_rtc_dt_match);
+
+static struct platform_driver ftrtc010_rtc_driver = {
+	.driver		= {
+		.name	= DRV_NAME,
+		.of_match_table = ftrtc010_rtc_dt_match,
+	},
+	.probe		= ftrtc010_rtc_probe,
+	.remove		= ftrtc010_rtc_remove,
+};
+
+module_platform_driver_probe(ftrtc010_rtc_driver, ftrtc010_rtc_probe);