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, 166 insertions, 0 deletions

diff --git a/drivers/acpi/thermal_lib.c b/drivers/acpi/thermal_lib.c
new file mode 100644
index 000000000..f81591927
--- /dev/null
+++ b/drivers/acpi/thermal_lib.c
@@ -0,0 +1,166 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright 2023 Linaro Limited
+ * Copyright 2023 Intel Corporation
+ *
+ * Library routines for retrieving trip point temperature values from the
+ * platform firmware via ACPI.
+ */
+#include <linux/acpi.h>
+#include <linux/units.h>
+#include <linux/thermal.h>
+#include "internal.h"
+
+/*
+ * Minimum temperature for full military grade is 218°K (-55°C) and
+ * max temperature is 448°K (175°C). We can consider those values as
+ * the boundaries for the [trips] temperature returned by the
+ * firmware. Any values out of these boundaries may be considered
+ * bogus and we can assume the firmware has no data to provide.
+ */
+#define TEMP_MIN_DECIK	2180ULL
+#define TEMP_MAX_DECIK	4480ULL
+
+static int acpi_trip_temp(struct acpi_device *adev, char *obj_name,
+			  int *ret_temp)
+{
+	unsigned long long temp;
+	acpi_status status;
+
+	status = acpi_evaluate_integer(adev->handle, obj_name, NULL, &temp);
+	if (ACPI_FAILURE(status)) {
+		acpi_handle_debug(adev->handle, "%s evaluation failed\n", obj_name);
+		return -ENODATA;
+	}
+
+	if (temp >= TEMP_MIN_DECIK && temp <= TEMP_MAX_DECIK) {
+		*ret_temp = temp;
+	} else {
+		acpi_handle_debug(adev->handle, "%s result %llu out of range\n",
+				  obj_name, temp);
+		*ret_temp = THERMAL_TEMP_INVALID;
+	}
+
+	return 0;
+}
+
+int acpi_active_trip_temp(struct acpi_device *adev, int id, int *ret_temp)
+{
+	char obj_name[] = {'_', 'A', 'C', '0' + id, '\0'};
+
+	if (id < 0 || id > 9)
+		return -EINVAL;
+
+	return acpi_trip_temp(adev, obj_name, ret_temp);
+}
+EXPORT_SYMBOL_NS_GPL(acpi_active_trip_temp, "ACPI_THERMAL");
+
+int acpi_passive_trip_temp(struct acpi_device *adev, int *ret_temp)
+{
+	return acpi_trip_temp(adev, "_PSV", ret_temp);
+}
+EXPORT_SYMBOL_NS_GPL(acpi_passive_trip_temp, "ACPI_THERMAL");
+
+int acpi_hot_trip_temp(struct acpi_device *adev, int *ret_temp)
+{
+	return acpi_trip_temp(adev, "_HOT", ret_temp);
+}
+EXPORT_SYMBOL_NS_GPL(acpi_hot_trip_temp, "ACPI_THERMAL");
+
+int acpi_critical_trip_temp(struct acpi_device *adev, int *ret_temp)
+{
+	return acpi_trip_temp(adev, "_CRT", ret_temp);
+}
+EXPORT_SYMBOL_NS_GPL(acpi_critical_trip_temp, "ACPI_THERMAL");
+
+static int thermal_temp(int error, int temp_decik, int *ret_temp)
+{
+	if (error)
+		return error;
+
+	if (temp_decik == THERMAL_TEMP_INVALID)
+		*ret_temp = THERMAL_TEMP_INVALID;
+	else
+		*ret_temp = deci_kelvin_to_millicelsius(temp_decik);
+
+	return 0;
+}
+
+/**
+ * thermal_acpi_active_trip_temp - Retrieve active trip point temperature
+ * @adev: Target thermal zone ACPI device object.
+ * @id: Active cooling level (0 - 9).
+ * @ret_temp: Address to store the retrieved temperature value on success.
+ *
+ * Evaluate the _ACx object for the thermal zone represented by @adev to obtain
+ * the temperature of the active cooling trip point corresponding to the active
+ * cooling level given by @id.
+ *
+ * Return 0 on success or a negative error value on failure.
+ */
+int thermal_acpi_active_trip_temp(struct acpi_device *adev, int id, int *ret_temp)
+{
+	int temp_decik = 0;
+	int ret = acpi_active_trip_temp(adev, id, &temp_decik);
+
+	return thermal_temp(ret, temp_decik, ret_temp);
+}
+EXPORT_SYMBOL_GPL(thermal_acpi_active_trip_temp);
+
+/**
+ * thermal_acpi_passive_trip_temp - Retrieve passive trip point temperature
+ * @adev: Target thermal zone ACPI device object.
+ * @ret_temp: Address to store the retrieved temperature value on success.
+ *
+ * Evaluate the _PSV object for the thermal zone represented by @adev to obtain
+ * the temperature of the passive cooling trip point.
+ *
+ * Return 0 on success or -ENODATA on failure.
+ */
+int thermal_acpi_passive_trip_temp(struct acpi_device *adev, int *ret_temp)
+{
+	int temp_decik = 0;
+	int ret = acpi_passive_trip_temp(adev, &temp_decik);
+
+	return thermal_temp(ret, temp_decik, ret_temp);
+}
+EXPORT_SYMBOL_GPL(thermal_acpi_passive_trip_temp);
+
+/**
+ * thermal_acpi_hot_trip_temp - Retrieve hot trip point temperature
+ * @adev: Target thermal zone ACPI device object.
+ * @ret_temp: Address to store the retrieved temperature value on success.
+ *
+ * Evaluate the _HOT object for the thermal zone represented by @adev to obtain
+ * the temperature of the trip point at which the system is expected to be put
+ * into the S4 sleep state.
+ *
+ * Return 0 on success or -ENODATA on failure.
+ */
+int thermal_acpi_hot_trip_temp(struct acpi_device *adev, int *ret_temp)
+{
+	int temp_decik = 0;
+	int ret = acpi_hot_trip_temp(adev, &temp_decik);
+
+	return thermal_temp(ret, temp_decik, ret_temp);
+}
+EXPORT_SYMBOL_GPL(thermal_acpi_hot_trip_temp);
+
+/**
+ * thermal_acpi_critical_trip_temp - Retrieve critical trip point temperature
+ * @adev: Target thermal zone ACPI device object.
+ * @ret_temp: Address to store the retrieved temperature value on success.
+ *
+ * Evaluate the _CRT object for the thermal zone represented by @adev to obtain
+ * the temperature of the critical cooling trip point.
+ *
+ * Return 0 on success or -ENODATA on failure.
+ */
+int thermal_acpi_critical_trip_temp(struct acpi_device *adev, int *ret_temp)
+{
+	int temp_decik = 0;
+	int ret = acpi_critical_trip_temp(adev, &temp_decik);
+
+	return thermal_temp(ret, temp_decik, ret_temp);
+}
+EXPORT_SYMBOL_GPL(thermal_acpi_critical_trip_temp);