thermal: add thermal trip event cooling device

This change adds a cooling device driver to notify the user space of the
thermal trip event. To avoid having user space process poll the
cooling state, a sysfs node is exposed that supports blocking reads.
The driver also supports a timeout (in milliseconds) for blocking reads
which can be done by writing the value to thermal_trip_event_block node
before reading. The blocked user space process will be woken up when the
cooling device becomes active or times out.

Bug 4261645
Bug 1688327

Signed-off-by: Yi-Wei Wang <yiweiw@nvidia.com>
Change-Id: Ic89406ba2713e5bc8f3806d6cfeb462601c73a7d
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nv-oot/+/3015652
(cherry picked from commit 856471d64f)
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nv-oot/+/3019439
Reviewed-by: Laxman Dewangan <ldewangan@nvidia.com>
GVS: Gerrit_Virtual_Submit <buildbot_gerritrpt@nvidia.com>

drivers/thermal/Makefile

@@ -4,5 +4,6 @@
 ifeq ($(findstring ack_src,$(NV_BUILD_KERNEL_OPTIONS)),)
 obj-m += pex9749-thermal.o
 obj-m += tegra234-oc-event.o
+obj-m += thermal-trip-event.o
 endif
 obj-m += max77851_thermal.o

drivers/thermal/thermal-trip-event.c

@@ -0,0 +1,257 @@
+// SPDX-License-Identifier: GPL-2.0-only
+// SPDX-FileCopyrightText: Copyright (c) 2023, NVIDIA CORPORATION & AFFILIATES.  All rights reserved.
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/platform_device.h>
+#include <linux/thermal.h>
+#include <linux/wait.h>
+
+enum cdev_states {
+	CDEV_INACTIVE,
+	CDEV_ACTIVE,
+	CDEV_DESTROY,
+};
+
+struct therm_trip_event {
+	unsigned int cur_state;
+	unsigned int max_state;
+	unsigned int event_timeout_ms;
+	struct mutex cur_state_lock;
+	struct mutex event_timeout_lock;
+	struct thermal_cooling_device *cdev;
+	wait_queue_head_t waitq_head;
+};
+
+static int tte_cdev_get_max_state(struct thermal_cooling_device *tcd,
+				  unsigned long *state)
+{
+	struct therm_trip_event *tte = tcd->devdata;
+
+	*state = tte->max_state;
+
+	return 0;
+}
+
+static int tte_cdev_get_cur_state(struct thermal_cooling_device *tcd,
+				  unsigned long *state)
+{
+	struct therm_trip_event *tte = tcd->devdata;
+
+	*state = tte->cur_state;
+
+	return 0;
+}
+
+static int tte_cdev_set_cur_state(struct thermal_cooling_device *tcd,
+				  unsigned long state)
+{
+	struct therm_trip_event *tte = tcd->devdata;
+	struct device *dev = &tcd->device;
+
+	if (state > tte->max_state)
+		return -EINVAL;
+
+	if (state == tte->cur_state)
+		return 0;
+
+	dev_notice(dev, "%s cooling state: %u -> %lu\n", tcd->type,
+		   tte->cur_state, state);
+
+	/*
+	 * Although tcd->lock is already held in the thermal framework, a lock
+	 * is added here to avoid a race condition between set_cur_state() and
+	 * driver removal.
+	 */
+	mutex_lock(&tte->cur_state_lock);
+	tte->cur_state = state;
+	mutex_unlock(&tte->cur_state_lock);
+
+	if (tte->cur_state != CDEV_INACTIVE) {
+		if (wq_has_sleeper(&tte->waitq_head)) {
+			wake_up_interruptible_all(&tte->waitq_head);
+			dev_dbg(dev, "THERMAL_EVENT_TRIPPED!\n");
+		}
+	}
+
+	return 0;
+}
+
+static ssize_t thermal_trip_event_show(struct device *dev,
+				       struct device_attribute *attr, char *buf)
+{
+	struct therm_trip_event *tte = dev_get_drvdata(dev);
+
+	return sprintf(buf, "%d\n", tte->cur_state);
+}
+
+static ssize_t thermal_trip_event_block_show(struct device *dev,
+					     struct device_attribute *attr,
+					     char *buf)
+{
+	struct therm_trip_event *tte = dev_get_drvdata(dev);
+	unsigned int event_timeout_ms = tte->event_timeout_ms;
+	int ret;
+
+	if (event_timeout_ms > 0)
+		ret = wait_event_interruptible_timeout(
+			tte->waitq_head, tte->cur_state != CDEV_INACTIVE,
+			msecs_to_jiffies(event_timeout_ms));
+	else
+		ret = wait_event_interruptible(tte->waitq_head,
+					       tte->cur_state != CDEV_INACTIVE);
+
+	/*
+	 * -ERESTARTSYS is returned to avoid false alarm and to resume the call.
+	 */
+	if (ret < 0)
+		return ret;
+
+	return sprintf(buf, "%d\n", tte->cur_state);
+}
+
+static ssize_t thermal_trip_event_block_store(struct device *dev,
+					      struct device_attribute *attr,
+					      const char *buf, size_t count)
+{
+	unsigned int val;
+	struct therm_trip_event *tte = dev_get_drvdata(dev);
+
+	if (kstrtouint(buf, 0, &val))
+		return -EINVAL;
+
+	mutex_lock(&tte->event_timeout_lock);
+	tte->event_timeout_ms = val;
+	mutex_unlock(&tte->event_timeout_lock);
+
+	return count;
+}
+
+static struct thermal_cooling_device_ops tte_cdev_ops = {
+	.get_max_state = tte_cdev_get_max_state,
+	.get_cur_state = tte_cdev_get_cur_state,
+	.set_cur_state = tte_cdev_set_cur_state,
+};
+
+static DEVICE_ATTR_RO(thermal_trip_event);
+static DEVICE_ATTR_RW(thermal_trip_event_block);
+
+static const struct attribute *tte_cdev_attr[] = {
+	&dev_attr_thermal_trip_event.attr,
+	&dev_attr_thermal_trip_event_block.attr,
+	NULL,
+};
+
+static int thermal_trip_event_probe(struct platform_device *pdev)
+{
+	struct device *dev = &pdev->dev;
+	struct device_node *np = dev->of_node;
+	struct therm_trip_event *tte;
+	const char *cdev_type;
+	int ret;
+
+	tte = devm_kzalloc(dev, sizeof(struct therm_trip_event), GFP_KERNEL);
+	if (!tte)
+		return -ENOMEM;
+
+	if (of_property_read_string(np, "cdev-type", &cdev_type) != 0) {
+		dev_err(dev, "invalid cdev-type property of %pOFn\n", np);
+		return -EINVAL;
+	}
+
+	tte->cur_state = CDEV_INACTIVE;
+	tte->max_state = CDEV_DESTROY;
+	mutex_init(&tte->cur_state_lock);
+	mutex_init(&tte->event_timeout_lock);
+	init_waitqueue_head(&tte->waitq_head);
+	dev_set_drvdata(dev, tte);
+
+	tte->cdev = thermal_of_cooling_device_register(np, cdev_type, tte,
+						       &tte_cdev_ops);
+	if (IS_ERR(tte->cdev)) {
+		ret = PTR_ERR(tte->cdev);
+		goto destroy_lock;
+	}
+
+	ret = sysfs_create_files(&dev->kobj, tte_cdev_attr);
+	if (ret) {
+		dev_err(dev, "failed to create sysfs files\n");
+		goto free_cdev;
+	}
+
+	ret = sysfs_create_link(&tte->cdev->device.kobj, &dev->kobj,
+				"thermal_trip_event");
+	if (ret) {
+		dev_err(dev, "failed to create sysfs symlink\n");
+		goto free_sysfs_files;
+	}
+
+	dev_info(dev, "cooling device registered.\n");
+	return 0;
+
+free_sysfs_files:
+	sysfs_remove_files(&dev->kobj, tte_cdev_attr);
+free_cdev:
+	thermal_cooling_device_unregister(tte->cdev);
+destroy_lock:
+	mutex_destroy(&tte->event_timeout_lock);
+	mutex_destroy(&tte->cur_state_lock);
+
+	return ret;
+}
+
+static int thermal_trip_event_remove(struct platform_device *pdev)
+{
+	struct device *dev = &pdev->dev;
+	struct therm_trip_event *tte = dev_get_drvdata(dev);
+	struct thermal_cooling_device *cdev = tte->cdev;
+
+	/*
+	 * There could be processes waiting for the event to happen while the
+	 * driver is being removed. To have smooth removal, wake them up by
+	 * updating the cur_state. The cooling state change here is not
+	 * perceived by the thermal framework, but it's not a big deal as the
+	 * cooling device is going to be destroyed soon.
+	 */
+	mutex_lock(&tte->cur_state_lock);
+	tte->cur_state = CDEV_DESTROY;
+	mutex_unlock(&tte->cur_state_lock);
+
+	if (wq_has_sleeper(&tte->waitq_head)) {
+		wake_up_interruptible_all(&tte->waitq_head);
+		dev_dbg(dev, "THERMAL_EVENT_DESTROYED!\n");
+	}
+
+	sysfs_remove_link(&cdev->device.kobj, "thermal_trip_event");
+	sysfs_remove_files(&dev->kobj, tte_cdev_attr);
+	thermal_cooling_device_unregister(cdev);
+	mutex_destroy(&tte->event_timeout_lock);
+	mutex_destroy(&tte->cur_state_lock);
+
+	return 0;
+}
+
+static const struct of_device_id thermal_trip_event_of_match[] = {
+	{
+		.compatible = "thermal-trip-event",
+	},
+	{},
+};
+MODULE_DEVICE_TABLE(of, thermal_trip_event_of_match);
+
+static struct platform_driver thermal_trip_event_driver = {
+	.driver = {
+		.name = "thermal-trip-event",
+		.owner = THIS_MODULE,
+		.of_match_table = thermal_trip_event_of_match,
+	},
+	.probe = thermal_trip_event_probe,
+	.remove = thermal_trip_event_remove,
+};
+
+module_platform_driver(thermal_trip_event_driver);
+
+MODULE_AUTHOR("Sreenivasulu Velpula <svelpula@nvidia.com>");
+MODULE_AUTHOR("Yi-Wei Wang <yiweiw@nvidia.com>");
+MODULE_DESCRIPTION("Thermal Trip Event Driver");
+MODULE_LICENSE("GPL v2");