Merge earlier changes in Intel thermal drivers for v6.7.

Documentation/driver-api/thermal/intel_dptf.rst

@@ -315,3 +315,57 @@ DPTF Fan Control
 ----------------------------------------
 
 Refer to Documentation/admin-guide/acpi/fan_performance_states.rst
+
+Workload Type Hints
+----------------------------------------
+
+The firmware in Meteor Lake processor generation is capable of identifying
+workload type and passing hints regarding it to the OS. A special sysfs
+interface is provided to allow user space to obtain workload type hints from
+the firmware and control the rate at which they are provided.
+
+User space can poll attribute "workload_type_index" for the current hint or
+can receive a notification whenever the value of this attribute is updated.
+
+file:`/sys/bus/pci/devices/0000:00:04.0/workload_hint/`
+Segment 0, bus 0, device 4, function 0 is reserved for the processor thermal
+device on all Intel client processors. So, the above path doesn't change
+based on the processor generation.
+
+``workload_hint_enable`` (RW)
+	Enable firmware to send workload type hints to user space.
+
+``notification_delay_ms`` (RW)
+	Minimum delay in milliseconds before firmware will notify OS. This is
+	for the rate control of notifications. This delay is between changing
+	the workload type prediction in the firmware and notifying the OS about
+	the change. The default delay is 1024 ms. The delay of 0 is invalid.
+	The delay is rounded up to the nearest power of 2 to simplify firmware
+	programming of the delay value. The read of notification_delay_ms
+	attribute shows the effective value used.
+
+``workload_type_index`` (RO)
+	Predicted workload type index. User space can get notification of
+	change via existing sysfs attribute change notification mechanism.
+
+	The supported index values and their meaning for the Meteor Lake
+	processor generation are as follows:
+
+	0 -  Idle: System performs no tasks, power and idle residency are
+		consistently low for long periods of time.
+
+	1 – Battery Life: Power is relatively low, but the processor may
+		still be actively performing a task, such as video playback for
+		a long period of time.
+
+	2 – Sustained: Power level that is relatively high for a long period
+		of time, with very few to no periods of idleness, which will
+		eventually exhaust RAPL Power Limit 1 and 2.
+
+	3 – Bursty: Consumes a relatively constant average amount of power, but
+		periods of relative idleness are interrupted by bursts of
+		activity. The bursts are relatively short and the periods of
+		relative idleness between them typically prevent RAPL Power
+		Limit 1 from being exhausted.
+
+	4 – Unknown: Can't classify.

drivers/thermal/intel/int340x_thermal/Makefile

@@ -10,5 +10,7 @@ obj-$(CONFIG_INT340X_THERMAL)	+= processor_thermal_device_pci.o
 obj-$(CONFIG_PROC_THERMAL_MMIO_RAPL) += processor_thermal_rapl.o
 obj-$(CONFIG_INT340X_THERMAL)	+= processor_thermal_rfim.o
 obj-$(CONFIG_INT340X_THERMAL)	+= processor_thermal_mbox.o
+obj-$(CONFIG_INT340X_THERMAL)	+= processor_thermal_wt_req.o
+obj-$(CONFIG_INT340X_THERMAL)	+= processor_thermal_wt_hint.o
 obj-$(CONFIG_INT3406_THERMAL)	+= int3406_thermal.o
 obj-$(CONFIG_ACPI_THERMAL_REL)	+= acpi_thermal_rel.o

drivers/thermal/intel/int340x_thermal/processor_thermal_device.c

@@ -346,12 +346,18 @@ int proc_thermal_mmio_add(struct pci_dev *pdev,
 		}
 	}
 
-	if (feature_mask & PROC_THERMAL_FEATURE_MBOX) {
-		ret = proc_thermal_mbox_add(pdev, proc_priv);
+	if (feature_mask & PROC_THERMAL_FEATURE_WT_REQ) {
+		ret = proc_thermal_wt_req_add(pdev, proc_priv);
 		if (ret) {
 			dev_err(&pdev->dev, "failed to add MBOX interface\n");
 			goto err_rem_rfim;
 		}
+	} else if (feature_mask & PROC_THERMAL_FEATURE_WT_HINT) {
+		ret = proc_thermal_wt_hint_add(pdev, proc_priv);
+		if (ret) {
+			dev_err(&pdev->dev, "failed to add WT Hint\n");
+			goto err_rem_rfim;
+		}
 	}
 
 	return 0;
@@ -374,12 +380,15 @@ void proc_thermal_mmio_remove(struct pci_dev *pdev, struct proc_thermal_device *
 	    proc_priv->mmio_feature_mask & PROC_THERMAL_FEATURE_DVFS)
 		proc_thermal_rfim_remove(pdev);
 
-	if (proc_priv->mmio_feature_mask & PROC_THERMAL_FEATURE_MBOX)
-		proc_thermal_mbox_remove(pdev);
+	if (proc_priv->mmio_feature_mask & PROC_THERMAL_FEATURE_WT_REQ)
+		proc_thermal_wt_req_remove(pdev);
+	else if (proc_priv->mmio_feature_mask & PROC_THERMAL_FEATURE_WT_HINT)
+		proc_thermal_wt_hint_remove(pdev);
 }
 EXPORT_SYMBOL_GPL(proc_thermal_mmio_remove);
 
 MODULE_IMPORT_NS(INTEL_TCC);
+MODULE_IMPORT_NS(INT340X_THERMAL);
 MODULE_AUTHOR("Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>");
 MODULE_DESCRIPTION("Processor Thermal Reporting Device Driver");
 MODULE_LICENSE("GPL v2");

drivers/thermal/intel/int340x_thermal/processor_thermal_device.h

@@ -10,6 +10,7 @@
 #include <linux/intel_rapl.h>
 
 #define PCI_DEVICE_ID_INTEL_ADL_THERMAL	0x461d
+#define PCI_DEVICE_ID_INTEL_ARL_S_THERMAL 0xAD03
 #define PCI_DEVICE_ID_INTEL_BDW_THERMAL	0x1603
 #define PCI_DEVICE_ID_INTEL_BSW_THERMAL	0x22DC
 
@@ -59,8 +60,9 @@ struct rapl_mmio_regs {
 #define PROC_THERMAL_FEATURE_RAPL	0x01
 #define PROC_THERMAL_FEATURE_FIVR	0x02
 #define PROC_THERMAL_FEATURE_DVFS	0x04
-#define PROC_THERMAL_FEATURE_MBOX	0x08
+#define PROC_THERMAL_FEATURE_WT_REQ	0x08
 #define PROC_THERMAL_FEATURE_DLVR	0x10
+#define PROC_THERMAL_FEATURE_WT_HINT	0x20
 
 #if IS_ENABLED(CONFIG_PROC_THERMAL_MMIO_RAPL)
 int proc_thermal_rapl_add(struct pci_dev *pdev, struct proc_thermal_device *proc_priv);
@@ -80,13 +82,27 @@ static void __maybe_unused proc_thermal_rapl_remove(void)
 int proc_thermal_rfim_add(struct pci_dev *pdev, struct proc_thermal_device *proc_priv);
 void proc_thermal_rfim_remove(struct pci_dev *pdev);
 
-int proc_thermal_mbox_add(struct pci_dev *pdev, struct proc_thermal_device *proc_priv);
-void proc_thermal_mbox_remove(struct pci_dev *pdev);
+int proc_thermal_wt_req_add(struct pci_dev *pdev, struct proc_thermal_device *proc_priv);
+void proc_thermal_wt_req_remove(struct pci_dev *pdev);
+
+#define MBOX_CMD_WORKLOAD_TYPE_READ	0x0E
+#define MBOX_CMD_WORKLOAD_TYPE_WRITE	0x0F
+
+#define MBOX_DATA_BIT_AC_DC		30
+#define MBOX_DATA_BIT_VALID		31
 
 int processor_thermal_send_mbox_read_cmd(struct pci_dev *pdev, u16 id, u64 *resp);
 int processor_thermal_send_mbox_write_cmd(struct pci_dev *pdev, u16 id, u32 data);
+int processor_thermal_mbox_interrupt_config(struct pci_dev *pdev, bool enable, int enable_bit,
+					    int time_window);
 int proc_thermal_add(struct device *dev, struct proc_thermal_device *priv);
 void proc_thermal_remove(struct proc_thermal_device *proc_priv);
+
+int proc_thermal_wt_hint_add(struct pci_dev *pdev, struct proc_thermal_device *proc_priv);
+void proc_thermal_wt_hint_remove(struct pci_dev *pdev);
+void proc_thermal_wt_intr_callback(struct pci_dev *pdev, struct proc_thermal_device *proc_priv);
+bool proc_thermal_check_wt_intr(struct proc_thermal_device *proc_priv);
+
 int proc_thermal_suspend(struct device *dev);
 int proc_thermal_resume(struct device *dev);
 int proc_thermal_mmio_add(struct pci_dev *pdev,

drivers/thermal/intel/int340x_thermal/processor_thermal_device_pci.c

@@ -15,6 +15,11 @@
 
 #define DRV_NAME "proc_thermal_pci"
 
+static bool use_msi;
+module_param(use_msi, bool, 0644);
+MODULE_PARM_DESC(use_msi,
+	"Use PCI MSI based interrupts for processor thermal device.");
+
 struct proc_thermal_pci {
 	struct pci_dev *pdev;
 	struct proc_thermal_device *proc_priv;
@@ -117,20 +122,44 @@ static void pkg_thermal_schedule_work(struct delayed_work *work)
 	schedule_delayed_work(work, ms);
 }
 
+static irqreturn_t proc_thermal_irq_thread_handler(int irq, void *devid)
+{
+	struct proc_thermal_pci *pci_info = devid;
+
+	proc_thermal_wt_intr_callback(pci_info->pdev, pci_info->proc_priv);
+
+	return IRQ_HANDLED;
+}
+
 static irqreturn_t proc_thermal_irq_handler(int irq, void *devid)
 {
 	struct proc_thermal_pci *pci_info = devid;
+	struct proc_thermal_device *proc_priv;
+	int ret = IRQ_HANDLED;
 	u32 status;
 
-	proc_thermal_mmio_read(pci_info, PROC_THERMAL_MMIO_INT_STATUS_0, &status);
+	proc_priv = pci_info->proc_priv;
 
+	if (proc_priv->mmio_feature_mask & PROC_THERMAL_FEATURE_WT_HINT) {
+		if (proc_thermal_check_wt_intr(pci_info->proc_priv))
+			ret = IRQ_WAKE_THREAD;
+	}
+
+	/*
+	 * Since now there are two sources of interrupts: one from thermal threshold
+	 * and another from workload hint, add a check if there was really a threshold
+	 * interrupt before scheduling work function for thermal threshold.
+	 */
+	proc_thermal_mmio_read(pci_info, PROC_THERMAL_MMIO_INT_STATUS_0, &status);
+	if (status) {
 		/* Disable enable interrupt flag */
 		proc_thermal_mmio_write(pci_info, PROC_THERMAL_MMIO_INT_ENABLE_0, 0);
-	pci_write_config_byte(pci_info->pdev, 0xdc, 0x01);
-
 		pkg_thermal_schedule_work(&pci_info->work);
+	}
 
-	return IRQ_HANDLED;
+	pci_write_config_byte(pci_info->pdev, 0xdc, 0x01);
+
+	return ret;
 }
 
 static int sys_get_curr_temp(struct thermal_zone_device *tzd, int *temp)
@@ -203,6 +232,7 @@ static int proc_thermal_pci_probe(struct pci_dev *pdev, const struct pci_device_
 	struct proc_thermal_device *proc_priv;
 	struct proc_thermal_pci *pci_info;
 	int irq_flag = 0, irq, ret;
+	bool msi_irq = false;
 
 	proc_priv = devm_kzalloc(&pdev->dev, sizeof(*proc_priv), GFP_KERNEL);
 	if (!proc_priv)
@@ -248,18 +278,23 @@ static int proc_thermal_pci_probe(struct pci_dev *pdev, const struct pci_device_
 		goto err_ret_mmio;
 	}
 
+	if (use_msi && (pdev->msi_enabled || pdev->msix_enabled)) {
 		/* request and enable interrupt */
 		ret = pci_alloc_irq_vectors(pdev, 1, 1, PCI_IRQ_ALL_TYPES);
 		if (ret < 0) {
 			dev_err(&pdev->dev, "Failed to allocate vectors!\n");
 			goto err_ret_tzone;
 		}
-	if (!pdev->msi_enabled && !pdev->msix_enabled)
-		irq_flag = IRQF_SHARED;
 
 		irq =  pci_irq_vector(pdev, 0);
+		msi_irq = true;
+	} else {
+		irq_flag = IRQF_SHARED;
+		irq = pdev->irq;
+	}
+
 	ret = devm_request_threaded_irq(&pdev->dev, irq,
-					proc_thermal_irq_handler, NULL,
+					proc_thermal_irq_handler, proc_thermal_irq_thread_handler,
 					irq_flag, KBUILD_MODNAME, pci_info);
 	if (ret) {
 		dev_err(&pdev->dev, "Request IRQ %d failed\n", pdev->irq);
@@ -273,6 +308,7 @@ static int proc_thermal_pci_probe(struct pci_dev *pdev, const struct pci_device_
 	return 0;
 
 err_free_vectors:
+	if (msi_irq)
 		pci_free_irq_vectors(pdev);
 err_ret_tzone:
 	thermal_zone_device_unregister(pci_info->tzone);
@@ -350,9 +386,15 @@ static SIMPLE_DEV_PM_OPS(proc_thermal_pci_pm, proc_thermal_pci_suspend,
 			 proc_thermal_pci_resume);
 
 static const struct pci_device_id proc_thermal_pci_ids[] = {
-	{ PCI_DEVICE_DATA(INTEL, ADL_THERMAL, PROC_THERMAL_FEATURE_RAPL | PROC_THERMAL_FEATURE_FIVR | PROC_THERMAL_FEATURE_DVFS | PROC_THERMAL_FEATURE_MBOX) },
-	{ PCI_DEVICE_DATA(INTEL, MTLP_THERMAL, PROC_THERMAL_FEATURE_RAPL | PROC_THERMAL_FEATURE_FIVR | PROC_THERMAL_FEATURE_DVFS | PROC_THERMAL_FEATURE_MBOX | PROC_THERMAL_FEATURE_DLVR) },
-	{ PCI_DEVICE_DATA(INTEL, RPL_THERMAL, PROC_THERMAL_FEATURE_RAPL | PROC_THERMAL_FEATURE_FIVR | PROC_THERMAL_FEATURE_DVFS | PROC_THERMAL_FEATURE_MBOX) },
+	{ PCI_DEVICE_DATA(INTEL, ADL_THERMAL, PROC_THERMAL_FEATURE_RAPL |
+	  PROC_THERMAL_FEATURE_FIVR | PROC_THERMAL_FEATURE_DVFS | PROC_THERMAL_FEATURE_WT_REQ) },
+	{ PCI_DEVICE_DATA(INTEL, MTLP_THERMAL, PROC_THERMAL_FEATURE_RAPL |
+	  PROC_THERMAL_FEATURE_FIVR | PROC_THERMAL_FEATURE_DVFS | PROC_THERMAL_FEATURE_DLVR |
+	  PROC_THERMAL_FEATURE_WT_HINT) },
+	{ PCI_DEVICE_DATA(INTEL, ARL_S_THERMAL, PROC_THERMAL_FEATURE_RAPL |
+	  PROC_THERMAL_FEATURE_DVFS | PROC_THERMAL_FEATURE_DLVR | PROC_THERMAL_FEATURE_WT_HINT) },
+	{ PCI_DEVICE_DATA(INTEL, RPL_THERMAL, PROC_THERMAL_FEATURE_RAPL |
+	  PROC_THERMAL_FEATURE_FIVR | PROC_THERMAL_FEATURE_DVFS | PROC_THERMAL_FEATURE_WT_REQ) },
 	{ },
 };
 
@@ -368,6 +410,8 @@ static struct pci_driver proc_thermal_pci_driver = {
 
 module_pci_driver(proc_thermal_pci_driver);
 
+MODULE_IMPORT_NS(INT340X_THERMAL);
+
 MODULE_AUTHOR("Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>");
 MODULE_DESCRIPTION("Processor Thermal Reporting Device Driver");
 MODULE_LICENSE("GPL v2");

drivers/thermal/intel/int340x_thermal/processor_thermal_device_pci_legacy.c

@@ -137,7 +137,8 @@ static const struct pci_device_id proc_thermal_pci_ids[] = {
 	{ PCI_DEVICE_DATA(INTEL, ICL_THERMAL, PROC_THERMAL_FEATURE_RAPL) },
 	{ PCI_DEVICE_DATA(INTEL, JSL_THERMAL, 0) },
 	{ PCI_DEVICE_DATA(INTEL, SKL_THERMAL, PROC_THERMAL_FEATURE_RAPL) },
-	{ PCI_DEVICE_DATA(INTEL, TGL_THERMAL, PROC_THERMAL_FEATURE_RAPL | PROC_THERMAL_FEATURE_FIVR | PROC_THERMAL_FEATURE_MBOX) },
+	{ PCI_DEVICE_DATA(INTEL, TGL_THERMAL, PROC_THERMAL_FEATURE_RAPL |
+	  PROC_THERMAL_FEATURE_FIVR | PROC_THERMAL_FEATURE_WT_REQ) },
 	{ },
 };
 

drivers/thermal/intel/int340x_thermal/processor_thermal_mbox.c

@@ -10,18 +10,12 @@
 #include <linux/io-64-nonatomic-lo-hi.h>
 #include "processor_thermal_device.h"
 
-#define MBOX_CMD_WORKLOAD_TYPE_READ	0x0E
-#define MBOX_CMD_WORKLOAD_TYPE_WRITE	0x0F
-
 #define MBOX_OFFSET_DATA		0x5810
 #define MBOX_OFFSET_INTERFACE		0x5818
 
 #define MBOX_BUSY_BIT			31
 #define MBOX_RETRY_COUNT		100
 
-#define MBOX_DATA_BIT_VALID		31
-#define MBOX_DATA_BIT_AC_DC		30
-
 static DEFINE_MUTEX(mbox_lock);
 
 static int wait_for_mbox_ready(struct proc_thermal_device *proc_priv)
@@ -51,23 +45,16 @@ static int send_mbox_write_cmd(struct pci_dev *pdev, u16 id, u32 data)
 	int ret;
 
 	proc_priv = pci_get_drvdata(pdev);
-
-	mutex_lock(&mbox_lock);
-
 	ret = wait_for_mbox_ready(proc_priv);
 	if (ret)
-		goto unlock_mbox;
+		return ret;
 
 	writel(data, (proc_priv->mmio_base + MBOX_OFFSET_DATA));
 	/* Write command register */
 	reg_data = BIT_ULL(MBOX_BUSY_BIT) | id;
 	writel(reg_data, (proc_priv->mmio_base + MBOX_OFFSET_INTERFACE));
 
-	ret = wait_for_mbox_ready(proc_priv);
-
-unlock_mbox:
-	mutex_unlock(&mbox_lock);
-	return ret;
+	return wait_for_mbox_ready(proc_priv);
 }
 
 static int send_mbox_read_cmd(struct pci_dev *pdev, u16 id, u64 *resp)
@@ -77,12 +64,9 @@ static int send_mbox_read_cmd(struct pci_dev *pdev, u16 id, u64 *resp)
 	int ret;
 
 	proc_priv = pci_get_drvdata(pdev);
-
-	mutex_lock(&mbox_lock);
-
 	ret = wait_for_mbox_ready(proc_priv);
 	if (ret)
-		goto unlock_mbox;
+		return ret;
 
 	/* Write command register */
 	reg_data = BIT_ULL(MBOX_BUSY_BIT) | id;
@@ -90,152 +74,85 @@ static int send_mbox_read_cmd(struct pci_dev *pdev, u16 id, u64 *resp)
 
 	ret = wait_for_mbox_ready(proc_priv);
 	if (ret)
-		goto unlock_mbox;
+		return ret;
 
 	if (id == MBOX_CMD_WORKLOAD_TYPE_READ)
 		*resp = readl(proc_priv->mmio_base + MBOX_OFFSET_DATA);
 	else
 		*resp = readq(proc_priv->mmio_base + MBOX_OFFSET_DATA);
 
-unlock_mbox:
-	mutex_unlock(&mbox_lock);
-	return ret;
+	return 0;
 }
 
 int processor_thermal_send_mbox_read_cmd(struct pci_dev *pdev, u16 id, u64 *resp)
 {
-	return send_mbox_read_cmd(pdev, id, resp);
-}
-EXPORT_SYMBOL_NS_GPL(processor_thermal_send_mbox_read_cmd, INT340X_THERMAL);
-
-int processor_thermal_send_mbox_write_cmd(struct pci_dev *pdev, u16 id, u32 data)
-{
-	return send_mbox_write_cmd(pdev, id, data);
-}
-EXPORT_SYMBOL_NS_GPL(processor_thermal_send_mbox_write_cmd, INT340X_THERMAL);
-
-/* List of workload types */
-static const char * const workload_types[] = {
-	"none",
-	"idle",
-	"semi_active",
-	"bursty",
-	"sustained",
-	"battery_life",
-	NULL
-};
-
-static ssize_t workload_available_types_show(struct device *dev,
-					       struct device_attribute *attr,
-					       char *buf)
-{
-	int i = 0;
-	int ret = 0;
-
-	while (workload_types[i] != NULL)
-		ret += sprintf(&buf[ret], "%s ", workload_types[i++]);
+	int ret;
 
-	ret += sprintf(&buf[ret], "\n");
+	mutex_lock(&mbox_lock);
+	ret = send_mbox_read_cmd(pdev, id, resp);
+	mutex_unlock(&mbox_lock);
 
 	return ret;
 }
+EXPORT_SYMBOL_NS_GPL(processor_thermal_send_mbox_read_cmd, INT340X_THERMAL);
 
-static DEVICE_ATTR_RO(workload_available_types);
-
-static ssize_t workload_type_store(struct device *dev,
-				    struct device_attribute *attr,
-				    const char *buf, size_t count)
+int processor_thermal_send_mbox_write_cmd(struct pci_dev *pdev, u16 id, u32 data)
 {
-	struct pci_dev *pdev = to_pci_dev(dev);
-	char str_preference[15];
-	u32 data = 0;
-	ssize_t ret;
+	int ret;
 
-	ret = sscanf(buf, "%14s", str_preference);
-	if (ret != 1)
-		return -EINVAL;
+	mutex_lock(&mbox_lock);
+	ret = send_mbox_write_cmd(pdev, id, data);
+	mutex_unlock(&mbox_lock);
 
-	ret = match_string(workload_types, -1, str_preference);
-	if (ret < 0)
 	return ret;
-
-	ret &= 0xff;
-
-	if (ret)
-		data = BIT(MBOX_DATA_BIT_VALID) | BIT(MBOX_DATA_BIT_AC_DC);
-
-	data |= ret;
-
-	ret = send_mbox_write_cmd(pdev, MBOX_CMD_WORKLOAD_TYPE_WRITE, data);
-	if (ret)
-		return false;
-
-	return count;
 }
+EXPORT_SYMBOL_NS_GPL(processor_thermal_send_mbox_write_cmd, INT340X_THERMAL);
 
-static ssize_t workload_type_show(struct device *dev,
-				   struct device_attribute *attr,
-				   char *buf)
+#define MBOX_CAMARILLO_RD_INTR_CONFIG	0x1E
+#define MBOX_CAMARILLO_WR_INTR_CONFIG	0x1F
+#define WLT_TW_MASK			GENMASK_ULL(30, 24)
+#define SOC_PREDICTION_TW_SHIFT		24
+
+int processor_thermal_mbox_interrupt_config(struct pci_dev *pdev, bool enable,
+					    int enable_bit, int time_window)
 {
-	struct pci_dev *pdev = to_pci_dev(dev);
-	u64 cmd_resp;
+	u64 data;
 	int ret;
 
-	ret = send_mbox_read_cmd(pdev, MBOX_CMD_WORKLOAD_TYPE_READ, &cmd_resp);
-	if (ret)
-		return false;
-
-	cmd_resp &= 0xff;
-
-	if (cmd_resp > ARRAY_SIZE(workload_types) - 1)
-		return -EINVAL;
-
-	return sprintf(buf, "%s\n", workload_types[cmd_resp]);
-}
+	if (!pdev)
+		return -ENODEV;
 
-static DEVICE_ATTR_RW(workload_type);
+	mutex_lock(&mbox_lock);
 
-static struct attribute *workload_req_attrs[] = {
-	&dev_attr_workload_available_types.attr,
-	&dev_attr_workload_type.attr,
-	NULL
-};
+	/* Do read modify write for MBOX_CAMARILLO_RD_INTR_CONFIG */
 
-static const struct attribute_group workload_req_attribute_group = {
-	.attrs = workload_req_attrs,
-	.name = "workload_request"
-};
+	ret = send_mbox_read_cmd(pdev, MBOX_CAMARILLO_RD_INTR_CONFIG,  &data);
+	if (ret) {
+		dev_err(&pdev->dev, "MBOX_CAMARILLO_RD_INTR_CONFIG failed\n");
+		goto unlock;
+	}
 
-static bool workload_req_created;
+	if (time_window >= 0) {
+		data &= ~WLT_TW_MASK;
 
-int proc_thermal_mbox_add(struct pci_dev *pdev, struct proc_thermal_device *proc_priv)
-{
-	u64 cmd_resp;
-	int ret;
+		/* Program notification delay */
+		data |= ((u64)time_window << SOC_PREDICTION_TW_SHIFT) & WLT_TW_MASK;
+	}
 
-	/* Check if there is a mailbox support, if fails return success */
-	ret = send_mbox_read_cmd(pdev, MBOX_CMD_WORKLOAD_TYPE_READ, &cmd_resp);
-	if (ret)
-		return 0;
+	if (enable)
+		data |= BIT(enable_bit);
+	else
+		data &= ~BIT(enable_bit);
 
-	ret = sysfs_create_group(&pdev->dev.kobj, &workload_req_attribute_group);
+	ret = send_mbox_write_cmd(pdev, MBOX_CAMARILLO_WR_INTR_CONFIG, data);
 	if (ret)
-		return ret;
+		dev_err(&pdev->dev, "MBOX_CAMARILLO_WR_INTR_CONFIG failed\n");
 
-	workload_req_created = true;
-
-	return 0;
-}
-EXPORT_SYMBOL_GPL(proc_thermal_mbox_add);
-
-void proc_thermal_mbox_remove(struct pci_dev *pdev)
-{
-	if (workload_req_created)
-		sysfs_remove_group(&pdev->dev.kobj, &workload_req_attribute_group);
-
-	workload_req_created = false;
+unlock:
+	mutex_unlock(&mbox_lock);
 
+	return ret;
 }
-EXPORT_SYMBOL_GPL(proc_thermal_mbox_remove);
+EXPORT_SYMBOL_NS_GPL(processor_thermal_mbox_interrupt_config, INT340X_THERMAL);
 
 MODULE_LICENSE("GPL v2");

drivers/thermal/intel/int340x_thermal/processor_thermal_wt_hint.c

+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * processor thermal device interface for reading workload type hints
+ * from the user space. The hints are provided by the firmware.
+ *
+ * Operation:
+ * When user space enables workload type prediction:
+ * - Use mailbox to:
+ *	Configure notification delay
+ *	Enable processor thermal device interrupt
+ *
+ * - The predicted workload type can be read from MMIO:
+ *	Offset 0x5B18 shows if there was an interrupt
+ *	active for change in workload type and also
+ *	predicted workload type.
+ *
+ * Two interface functions are provided to call when there is a
+ * thermal device interrupt:
+ * - proc_thermal_check_wt_intr():
+ *     Check if the interrupt is for change in workload type. Called from
+ *     interrupt context.
+ *
+ * - proc_thermal_wt_intr_callback():
+ *     Callback for interrupt processing in thread context. This involves
+ *	sending notification to user space that there is a change in the
+ *     workload type.
+ *
+ * Copyright (c) 2023, Intel Corporation.
+ */
+
+#include <linux/bitfield.h>
+#include <linux/pci.h>
+#include "processor_thermal_device.h"
+
+#define SOC_WT_RES_INT_STATUS_OFFSET	0x5B18
+#define SOC_WT				GENMASK_ULL(47, 40)
+
+#define SOC_WT_PREDICTION_INT_ENABLE_BIT	23
+
+#define SOC_WT_PREDICTION_INT_ACTIVE	BIT(2)
+
+/*
+ * Closest possible to 1 Second is 1024 ms with programmed time delay
+ * of 0x0A.
+ */
+static u8 notify_delay = 0x0A;
+static u16 notify_delay_ms = 1024;
+
+static DEFINE_MUTEX(wt_lock);
+static u8 wt_enable;
+
+/* Show current predicted workload type index */
+static ssize_t workload_type_index_show(struct device *dev,
+					struct device_attribute *attr,
+					char *buf)
+{
+	struct proc_thermal_device *proc_priv;
+	struct pci_dev *pdev = to_pci_dev(dev);
+	u64 status = 0;
+	int wt;
+
+	mutex_lock(&wt_lock);
+	if (!wt_enable) {
+		mutex_unlock(&wt_lock);
+		return -ENODATA;
+	}
+
+	proc_priv = pci_get_drvdata(pdev);
+
+	status = readq(proc_priv->mmio_base + SOC_WT_RES_INT_STATUS_OFFSET);
+
+	mutex_unlock(&wt_lock);
+
+	wt = FIELD_GET(SOC_WT, status);
+
+	return sysfs_emit(buf, "%d\n", wt);
+}
+
+static DEVICE_ATTR_RO(workload_type_index);
+
+static ssize_t workload_hint_enable_show(struct device *dev,
+					 struct device_attribute *attr,
+					 char *buf)
+{
+	return sysfs_emit(buf, "%d\n", wt_enable);
+}
+
+static ssize_t workload_hint_enable_store(struct device *dev,
+					  struct device_attribute *attr,
+					  const char *buf, size_t size)
+{
+	struct pci_dev *pdev = to_pci_dev(dev);
+	u8 mode;
+	int ret;
+
+	if (kstrtou8(buf, 10, &mode) || mode > 1)
+		return -EINVAL;
+
+	mutex_lock(&wt_lock);
+
+	if (mode)
+		ret = processor_thermal_mbox_interrupt_config(pdev, true,
+							      SOC_WT_PREDICTION_INT_ENABLE_BIT,
+							      notify_delay);
+	else
+		ret = processor_thermal_mbox_interrupt_config(pdev, false,
+							      SOC_WT_PREDICTION_INT_ENABLE_BIT, 0);
+
+	if (ret)
+		goto ret_enable_store;
+
+	ret = size;
+	wt_enable = mode;
+
+ret_enable_store:
+	mutex_unlock(&wt_lock);
+
+	return ret;
+}
+
+static DEVICE_ATTR_RW(workload_hint_enable);
+
+static ssize_t notification_delay_ms_show(struct device *dev,
+					  struct device_attribute *attr,
+					  char *buf)
+{
+	return sysfs_emit(buf, "%u\n", notify_delay_ms);
+}
+
+static ssize_t notification_delay_ms_store(struct device *dev,
+					   struct device_attribute *attr,
+					   const char *buf, size_t size)
+{
+	struct pci_dev *pdev = to_pci_dev(dev);
+	u16 new_tw;
+	int ret;
+	u8 tm;
+
+	/*
+	 * Time window register value:
+	 * Formula: (1 + x/4) * power(2,y)
+	 * x = 2 msbs, that is [30:29] y = 5 [28:24]
+	 * in INTR_CONFIG register.
+	 * The result will be in milli seconds.
+	 * Here, just keep x = 0, and just change y.
+	 * First round up the user value to power of 2 and
+	 * then take log2, to get "y" value to program.
+	 */
+	ret = kstrtou16(buf, 10, &new_tw);
+	if (ret)
+		return ret;
+
+	if (!new_tw)
+		return -EINVAL;
+
+	new_tw = roundup_pow_of_two(new_tw);
+	tm = ilog2(new_tw);
+	if (tm > 31)
+		return -EINVAL;
+
+	mutex_lock(&wt_lock);
+
+	/* If the workload hint was already enabled, then update with the new delay */
+	if (wt_enable)
+		ret = processor_thermal_mbox_interrupt_config(pdev, true,
+							      SOC_WT_PREDICTION_INT_ENABLE_BIT,
+							      tm);
+
+	if (!ret) {
+		ret = size;
+		notify_delay = tm;
+		notify_delay_ms = new_tw;
+	}
+
+	mutex_unlock(&wt_lock);
+
+	return ret;
+}
+
+static DEVICE_ATTR_RW(notification_delay_ms);
+
+static struct attribute *workload_hint_attrs[] = {
+	&dev_attr_workload_type_index.attr,
+	&dev_attr_workload_hint_enable.attr,
+	&dev_attr_notification_delay_ms.attr,
+	NULL
+};
+
+static const struct attribute_group workload_hint_attribute_group = {
+	.attrs = workload_hint_attrs,
+	.name = "workload_hint"
+};
+
+/*
+ * Callback to check if the interrupt for prediction is active.
+ * Caution: Called from the interrupt context.
+ */
+bool proc_thermal_check_wt_intr(struct proc_thermal_device *proc_priv)
+{
+	u64 int_status;
+
+	int_status = readq(proc_priv->mmio_base + SOC_WT_RES_INT_STATUS_OFFSET);
+	if (int_status & SOC_WT_PREDICTION_INT_ACTIVE)
+		return true;
+
+	return false;
+}
+EXPORT_SYMBOL_NS_GPL(proc_thermal_check_wt_intr, INT340X_THERMAL);
+
+/* Callback to notify user space */
+void proc_thermal_wt_intr_callback(struct pci_dev *pdev, struct proc_thermal_device *proc_priv)
+{
+	u64 status;
+
+	status = readq(proc_priv->mmio_base + SOC_WT_RES_INT_STATUS_OFFSET);
+	if (!(status & SOC_WT_PREDICTION_INT_ACTIVE))
+		return;
+
+	writeq(status & ~SOC_WT_PREDICTION_INT_ACTIVE,
+		       proc_priv->mmio_base + SOC_WT_RES_INT_STATUS_OFFSET);
+	sysfs_notify(&pdev->dev.kobj, "workload_hint", "workload_type_index");
+}
+EXPORT_SYMBOL_NS_GPL(proc_thermal_wt_intr_callback, INT340X_THERMAL);
+
+static bool workload_hint_created;
+
+int proc_thermal_wt_hint_add(struct pci_dev *pdev, struct proc_thermal_device *proc_priv)
+{
+	int ret;
+
+	ret = sysfs_create_group(&pdev->dev.kobj, &workload_hint_attribute_group);
+	if (ret)
+		return ret;
+
+	workload_hint_created = true;
+
+	return 0;
+}
+EXPORT_SYMBOL_NS_GPL(proc_thermal_wt_hint_add, INT340X_THERMAL);
+
+void proc_thermal_wt_hint_remove(struct pci_dev *pdev)
+{
+	mutex_lock(&wt_lock);
+	if (wt_enable)
+		processor_thermal_mbox_interrupt_config(pdev, false,
+							SOC_WT_PREDICTION_INT_ENABLE_BIT,
+							0);
+	mutex_unlock(&wt_lock);
+
+	if (workload_hint_created)
+		sysfs_remove_group(&pdev->dev.kobj, &workload_hint_attribute_group);
+
+	workload_hint_created = false;
+}
+EXPORT_SYMBOL_NS_GPL(proc_thermal_wt_hint_remove, INT340X_THERMAL);
+
+MODULE_IMPORT_NS(INT340X_THERMAL);
+MODULE_LICENSE("GPL");

drivers/thermal/intel/int340x_thermal/processor_thermal_wt_req.c

+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * processor thermal device for Workload type hints
+ * update from user space
+ *
+ * Copyright (c) 2020-2023, Intel Corporation.
+ */
+
+#include <linux/pci.h>
+#include "processor_thermal_device.h"
+
+/* List of workload types */
+static const char * const workload_types[] = {
+	"none",
+	"idle",
+	"semi_active",
+	"bursty",
+	"sustained",
+	"battery_life",
+	NULL
+};
+
+static ssize_t workload_available_types_show(struct device *dev,
+					     struct device_attribute *attr,
+					     char *buf)
+{
+	int i = 0;
+	int ret = 0;
+
+	while (workload_types[i] != NULL)
+		ret += sprintf(&buf[ret], "%s ", workload_types[i++]);
+
+	ret += sprintf(&buf[ret], "\n");
+
+	return ret;
+}
+
+static DEVICE_ATTR_RO(workload_available_types);
+
+static ssize_t workload_type_store(struct device *dev,
+				   struct device_attribute *attr,
+				   const char *buf, size_t count)
+{
+	struct pci_dev *pdev = to_pci_dev(dev);
+	char str_preference[15];
+	u32 data = 0;
+	ssize_t ret;
+
+	ret = sscanf(buf, "%14s", str_preference);
+	if (ret != 1)
+		return -EINVAL;
+
+	ret = match_string(workload_types, -1, str_preference);
+	if (ret < 0)
+		return ret;
+
+	ret &= 0xff;
+
+	if (ret)
+		data = BIT(MBOX_DATA_BIT_VALID) | BIT(MBOX_DATA_BIT_AC_DC);
+
+	data |= ret;
+
+	ret = processor_thermal_send_mbox_write_cmd(pdev, MBOX_CMD_WORKLOAD_TYPE_WRITE, data);
+	if (ret)
+		return false;
+
+	return count;
+}
+
+static ssize_t workload_type_show(struct device *dev,
+				  struct device_attribute *attr,
+				  char *buf)
+{
+	struct pci_dev *pdev = to_pci_dev(dev);
+	u64 cmd_resp;
+	int ret;
+
+	ret = processor_thermal_send_mbox_read_cmd(pdev, MBOX_CMD_WORKLOAD_TYPE_READ, &cmd_resp);
+	if (ret)
+		return false;
+
+	cmd_resp &= 0xff;
+
+	if (cmd_resp > ARRAY_SIZE(workload_types) - 1)
+		return -EINVAL;
+
+	return sprintf(buf, "%s\n", workload_types[cmd_resp]);
+}
+
+static DEVICE_ATTR_RW(workload_type);
+
+static struct attribute *workload_req_attrs[] = {
+	&dev_attr_workload_available_types.attr,
+	&dev_attr_workload_type.attr,
+	NULL
+};
+
+static const struct attribute_group workload_req_attribute_group = {
+	.attrs = workload_req_attrs,
+	.name = "workload_request"
+};
+
+static bool workload_req_created;
+
+int proc_thermal_wt_req_add(struct pci_dev *pdev, struct proc_thermal_device *proc_priv)
+{
+	u64 cmd_resp;
+	int ret;
+
+	/* Check if there is a mailbox support, if fails return success */
+	ret = processor_thermal_send_mbox_read_cmd(pdev, MBOX_CMD_WORKLOAD_TYPE_READ, &cmd_resp);
+	if (ret)
+		return 0;
+
+	ret = sysfs_create_group(&pdev->dev.kobj, &workload_req_attribute_group);
+	if (ret)
+		return ret;
+
+	workload_req_created = true;
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(proc_thermal_wt_req_add);
+
+void proc_thermal_wt_req_remove(struct pci_dev *pdev)
+{
+	if (workload_req_created)
+		sysfs_remove_group(&pdev->dev.kobj, &workload_req_attribute_group);
+
+	workload_req_created = false;
+}
+EXPORT_SYMBOL_GPL(proc_thermal_wt_req_remove);
+
+MODULE_IMPORT_NS(INT340X_THERMAL);
+MODULE_LICENSE("GPL");

tools/testing/selftests/Makefile

@@ -85,6 +85,7 @@ TARGETS += syscall_user_dispatch
 TARGETS += sysctl
 TARGETS += tc-testing
 TARGETS += tdx
+TARGETS += thermal/intel/workload_hint
 TARGETS += timens
 ifneq (1, $(quicktest))
 TARGETS += timers

tools/testing/selftests/thermal/intel/workload_hint/Makefile

+# SPDX-License-Identifier: GPL-2.0
+ifndef CROSS_COMPILE
+uname_M := $(shell uname -m 2>/dev/null || echo not)
+ARCH ?= $(shell echo $(uname_M) | sed -e s/i.86/x86/ -e s/x86_64/x86/)
+
+ifeq ($(ARCH),x86)
+TEST_GEN_PROGS := workload_hint_test
+
+include ../../../lib.mk
+
+endif
+endif

tools/testing/selftests/thermal/intel/workload_hint/workload_hint_test.c

+// SPDX-License-Identifier: GPL-2.0
+
+#define _GNU_SOURCE
+
+#include <stdio.h>
+#include <string.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <fcntl.h>
+#include <poll.h>
+#include <signal.h>
+
+#define WORKLOAD_NOTIFICATION_DELAY_ATTRIBUTE "/sys/bus/pci/devices/0000:00:04.0/workload_hint/notification_delay_ms"
+#define WORKLOAD_ENABLE_ATTRIBUTE "/sys/bus/pci/devices/0000:00:04.0/workload_hint/workload_hint_enable"
+#define WORKLOAD_TYPE_INDEX_ATTRIBUTE  "/sys/bus/pci/devices/0000:00:04.0/workload_hint/workload_type_index"
+
+static const char * const workload_types[] = {
+	"idle",
+	"battery_life",
+	"sustained",
+	"bursty",
+	NULL
+};
+
+#define WORKLOAD_TYPE_MAX_INDEX	3
+
+void workload_hint_exit(int signum)
+{
+	int fd;
+
+	/* Disable feature via sysfs knob */
+
+	fd = open(WORKLOAD_ENABLE_ATTRIBUTE, O_RDWR);
+	if (fd < 0) {
+		perror("Unable to open workload type feature enable file\n");
+		exit(1);
+	}
+
+	if (write(fd, "0\n", 2) < 0) {
+		perror("Can' disable workload hints\n");
+		exit(1);
+	}
+
+	printf("Disabled workload type prediction\n");
+
+	close(fd);
+}
+
+int main(int argc, char **argv)
+{
+	struct pollfd ufd;
+	char index_str[4];
+	int fd, ret, index;
+	char delay_str[64];
+	int delay = 0;
+
+	printf("Usage: workload_hint_test [notification delay in milli seconds]\n");
+
+	if (argc > 1) {
+		ret = sscanf(argv[1], "%d", &delay);
+		if (ret < 0) {
+			printf("Invalid delay\n");
+			exit(1);
+		}
+
+		printf("Setting notification delay to %d ms\n", delay);
+		if (delay < 0)
+			exit(1);
+
+		sprintf(delay_str, "%s\n", argv[1]);
+
+		sprintf(delay_str, "%s\n", argv[1]);
+		fd = open(WORKLOAD_NOTIFICATION_DELAY_ATTRIBUTE, O_RDWR);
+		if (fd < 0) {
+			perror("Unable to open workload notification delay\n");
+			exit(1);
+		}
+
+		if (write(fd, delay_str, strlen(delay_str)) < 0) {
+			perror("Can't set delay\n");
+			exit(1);
+		}
+
+		close(fd);
+	}
+
+	if (signal(SIGINT, workload_hint_exit) == SIG_IGN)
+		signal(SIGINT, SIG_IGN);
+	if (signal(SIGHUP, workload_hint_exit) == SIG_IGN)
+		signal(SIGHUP, SIG_IGN);
+	if (signal(SIGTERM, workload_hint_exit) == SIG_IGN)
+		signal(SIGTERM, SIG_IGN);
+
+	/* Enable feature via sysfs knob */
+	fd = open(WORKLOAD_ENABLE_ATTRIBUTE, O_RDWR);
+	if (fd < 0) {
+		perror("Unable to open workload type feature enable file\n");
+		exit(1);
+	}
+
+	if (write(fd, "1\n", 2) < 0) {
+		perror("Can' enable workload hints\n");
+		exit(1);
+	}
+
+	close(fd);
+
+	printf("Enabled workload type prediction\n");
+
+	while (1) {
+		fd = open(WORKLOAD_TYPE_INDEX_ATTRIBUTE, O_RDONLY);
+		if (fd < 0) {
+			perror("Unable to open workload type file\n");
+			exit(1);
+		}
+
+		if ((lseek(fd, 0L, SEEK_SET)) < 0) {
+			fprintf(stderr, "Failed to set pointer to beginning\n");
+			exit(1);
+		}
+
+		if (read(fd, index_str, sizeof(index_str)) < 0) {
+			fprintf(stderr, "Failed to read from:%s\n",
+			WORKLOAD_TYPE_INDEX_ATTRIBUTE);
+			exit(1);
+		}
+
+		ufd.fd = fd;
+		ufd.events = POLLPRI;
+
+		ret = poll(&ufd, 1, -1);
+		if (ret < 0) {
+			perror("poll error");
+			exit(1);
+		} else if (ret == 0) {
+			printf("Poll Timeout\n");
+		} else {
+			if ((lseek(fd, 0L, SEEK_SET)) < 0) {
+				fprintf(stderr, "Failed to set pointer to beginning\n");
+				exit(1);
+			}
+
+			if (read(fd, index_str, sizeof(index_str)) < 0)
+				exit(0);
+
+			ret = sscanf(index_str, "%d", &index);
+			if (ret < 0)
+				break;
+			if (index > WORKLOAD_TYPE_MAX_INDEX)
+				printf("Invalid workload type index\n");
+			else
+				printf("workload type:%s\n", workload_types[index]);
+		}
+
+		close(fd);
+	}
+}