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, 235 insertions, 0 deletions

diff --git a/drivers/hwmon/surface_temp.c b/drivers/hwmon/surface_temp.c
new file mode 100644
index 000000000..cd21f331f
--- /dev/null
+++ b/drivers/hwmon/surface_temp.c
@@ -0,0 +1,235 @@
+// SPDX-License-Identifier: GPL-2.0+
+/*
+ * Thermal sensor subsystem driver for Surface System Aggregator Module (SSAM).
+ *
+ * Copyright (C) 2022-2023 Maximilian Luz <luzmaximilian@gmail.com>
+ */
+
+#include <linux/bitops.h>
+#include <linux/hwmon.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/types.h>
+
+#include <linux/surface_aggregator/controller.h>
+#include <linux/surface_aggregator/device.h>
+
+/* -- SAM interface. -------------------------------------------------------- */
+
+/*
+ * Available sensors are indicated by a 16-bit bitfield, where a 1 marks the
+ * presence of a sensor. So we have at most 16 possible sensors/channels.
+ */
+#define SSAM_TMP_SENSOR_MAX_COUNT	16
+
+/*
+ * All names observed so far are 6 characters long, but there's only
+ * zeros after the name, so perhaps they can be longer. This number reflects
+ * the maximum zero-padded space observed in the returned buffer.
+ */
+#define SSAM_TMP_SENSOR_NAME_LENGTH	18
+
+struct ssam_tmp_get_name_rsp {
+	__le16 unknown1;
+	char unknown2;
+	char name[SSAM_TMP_SENSOR_NAME_LENGTH];
+} __packed;
+
+static_assert(sizeof(struct ssam_tmp_get_name_rsp) == 21);
+
+SSAM_DEFINE_SYNC_REQUEST_CL_R(__ssam_tmp_get_available_sensors, __le16, {
+	.target_category = SSAM_SSH_TC_TMP,
+	.command_id      = 0x04,
+});
+
+SSAM_DEFINE_SYNC_REQUEST_MD_R(__ssam_tmp_get_temperature, __le16, {
+	.target_category = SSAM_SSH_TC_TMP,
+	.command_id      = 0x01,
+});
+
+SSAM_DEFINE_SYNC_REQUEST_MD_R(__ssam_tmp_get_name, struct ssam_tmp_get_name_rsp, {
+	.target_category = SSAM_SSH_TC_TMP,
+	.command_id      = 0x0e,
+});
+
+static int ssam_tmp_get_available_sensors(struct ssam_device *sdev, s16 *sensors)
+{
+	__le16 sensors_le;
+	int status;
+
+	status = __ssam_tmp_get_available_sensors(sdev, &sensors_le);
+	if (status)
+		return status;
+
+	*sensors = le16_to_cpu(sensors_le);
+	return 0;
+}
+
+static int ssam_tmp_get_temperature(struct ssam_device *sdev, u8 iid, long *temperature)
+{
+	__le16 temp_le;
+	int status;
+
+	status = __ssam_tmp_get_temperature(sdev->ctrl, sdev->uid.target, iid, &temp_le);
+	if (status)
+		return status;
+
+	/* Convert 1/10 °K to 1/1000 °C */
+	*temperature = (le16_to_cpu(temp_le) - 2731) * 100L;
+	return 0;
+}
+
+static int ssam_tmp_get_name(struct ssam_device *sdev, u8 iid, char *buf, size_t buf_len)
+{
+	struct ssam_tmp_get_name_rsp name_rsp;
+	int status;
+
+	status =  __ssam_tmp_get_name(sdev->ctrl, sdev->uid.target, iid, &name_rsp);
+	if (status)
+		return status;
+
+	/*
+	 * This should not fail unless the name in the returned struct is not
+	 * null-terminated or someone changed something in the struct
+	 * definitions above, since our buffer and struct have the same
+	 * capacity by design. So if this fails, log an error message. Since
+	 * the more likely cause is that the returned string isn't
+	 * null-terminated, we might have received garbage (as opposed to just
+	 * an incomplete string), so also fail the function.
+	 */
+	status = strscpy(buf, name_rsp.name, buf_len);
+	if (status < 0) {
+		dev_err(&sdev->dev, "received non-null-terminated sensor name string\n");
+		return status;
+	}
+
+	return 0;
+}
+
+/* -- Driver.---------------------------------------------------------------- */
+
+struct ssam_temp {
+	struct ssam_device *sdev;
+	s16 sensors;
+	char names[SSAM_TMP_SENSOR_MAX_COUNT][SSAM_TMP_SENSOR_NAME_LENGTH];
+};
+
+static umode_t ssam_temp_hwmon_is_visible(const void *data,
+					  enum hwmon_sensor_types type,
+					  u32 attr, int channel)
+{
+	const struct ssam_temp *ssam_temp = data;
+
+	if (!(ssam_temp->sensors & BIT(channel)))
+		return 0;
+
+	return 0444;
+}
+
+static int ssam_temp_hwmon_read(struct device *dev,
+				enum hwmon_sensor_types type,
+				u32 attr, int channel, long *value)
+{
+	const struct ssam_temp *ssam_temp = dev_get_drvdata(dev);
+
+	return ssam_tmp_get_temperature(ssam_temp->sdev, channel + 1, value);
+}
+
+static int ssam_temp_hwmon_read_string(struct device *dev,
+				       enum hwmon_sensor_types type,
+				       u32 attr, int channel, const char **str)
+{
+	const struct ssam_temp *ssam_temp = dev_get_drvdata(dev);
+
+	*str = ssam_temp->names[channel];
+	return 0;
+}
+
+static const struct hwmon_channel_info * const ssam_temp_hwmon_info[] = {
+	HWMON_CHANNEL_INFO(chip,
+			   HWMON_C_REGISTER_TZ),
+	HWMON_CHANNEL_INFO(temp,
+			   HWMON_T_INPUT | HWMON_T_LABEL,
+			   HWMON_T_INPUT | HWMON_T_LABEL,
+			   HWMON_T_INPUT | HWMON_T_LABEL,
+			   HWMON_T_INPUT | HWMON_T_LABEL,
+			   HWMON_T_INPUT | HWMON_T_LABEL,
+			   HWMON_T_INPUT | HWMON_T_LABEL,
+			   HWMON_T_INPUT | HWMON_T_LABEL,
+			   HWMON_T_INPUT | HWMON_T_LABEL,
+			   HWMON_T_INPUT | HWMON_T_LABEL,
+			   HWMON_T_INPUT | HWMON_T_LABEL,
+			   HWMON_T_INPUT | HWMON_T_LABEL,
+			   HWMON_T_INPUT | HWMON_T_LABEL,
+			   HWMON_T_INPUT | HWMON_T_LABEL,
+			   HWMON_T_INPUT | HWMON_T_LABEL,
+			   HWMON_T_INPUT | HWMON_T_LABEL,
+			   HWMON_T_INPUT | HWMON_T_LABEL),
+	NULL
+};
+
+static const struct hwmon_ops ssam_temp_hwmon_ops = {
+	.is_visible = ssam_temp_hwmon_is_visible,
+	.read = ssam_temp_hwmon_read,
+	.read_string = ssam_temp_hwmon_read_string,
+};
+
+static const struct hwmon_chip_info ssam_temp_hwmon_chip_info = {
+	.ops = &ssam_temp_hwmon_ops,
+	.info = ssam_temp_hwmon_info,
+};
+
+static int ssam_temp_probe(struct ssam_device *sdev)
+{
+	struct ssam_temp *ssam_temp;
+	struct device *hwmon_dev;
+	s16 sensors;
+	int channel;
+	int status;
+
+	status = ssam_tmp_get_available_sensors(sdev, &sensors);
+	if (status)
+		return status;
+
+	ssam_temp = devm_kzalloc(&sdev->dev, sizeof(*ssam_temp), GFP_KERNEL);
+	if (!ssam_temp)
+		return -ENOMEM;
+
+	ssam_temp->sdev = sdev;
+	ssam_temp->sensors = sensors;
+
+	/* Retrieve the name for each available sensor. */
+	for (channel = 0; channel < SSAM_TMP_SENSOR_MAX_COUNT; channel++) {
+		if (!(sensors & BIT(channel)))
+			continue;
+
+		status = ssam_tmp_get_name(sdev, channel + 1, ssam_temp->names[channel],
+					   SSAM_TMP_SENSOR_NAME_LENGTH);
+		if (status)
+			return status;
+	}
+
+	hwmon_dev = devm_hwmon_device_register_with_info(&sdev->dev, "surface_thermal", ssam_temp,
+							 &ssam_temp_hwmon_chip_info, NULL);
+	return PTR_ERR_OR_ZERO(hwmon_dev);
+}
+
+static const struct ssam_device_id ssam_temp_match[] = {
+	{ SSAM_SDEV(TMP, SAM, 0x00, 0x02) },
+	{ },
+};
+MODULE_DEVICE_TABLE(ssam, ssam_temp_match);
+
+static struct ssam_device_driver ssam_temp = {
+	.probe = ssam_temp_probe,
+	.match_table = ssam_temp_match,
+	.driver = {
+		.name = "surface_temp",
+		.probe_type = PROBE_PREFER_ASYNCHRONOUS,
+	},
+};
+module_ssam_device_driver(ssam_temp);
+
+MODULE_AUTHOR("Maximilian Luz <luzmaximilian@gmail.com>");
+MODULE_DESCRIPTION("Thermal sensor subsystem driver for Surface System Aggregator Module");
+MODULE_LICENSE("GPL");