Re: [PATCH 2/5] thermal: int340x: processor_thermal: Support power floor notifications

From: Rafael J. Wysocki
Date: Mon Oct 02 2023 - 14:42:07 EST


On Wed, Sep 27, 2023 at 12:58 AM Srinivas Pandruvada
<srinivas.pandruvada@xxxxxxxxxxxxxxx> wrote:
>
> When the hardware reduces the power to the minimum possible, the power
> floor is notified via an interrupt. This can happen when user space
> requests a power limit via powercap RAPL interface, which forces the
> system to enter to the lowest power. This power floor indication can
> be used as a hint to resort to other methods of reducing power than
> via RAPL power limit.
>
> Before power floor status can be read or get notifications from the
> firmware, it needs to be configured via a mailbox command. Actual power
> floor status is read via bit 39 of MMIO offset 0x5B18 of the processor
> thermal PCI device.
>
> To show the current power floor status and get notification
> on a sysfs attribute, add additional attributes to
> /sys/bus/pci/devices/0000\:00\:04.0/power_limits/
>
> power_floor_enable : This attribute is present when a SoC supports
> power floor notification. This attribute allows to enable/disable
> power floor notifications.
>
> power_floor_status : This attribute is present when a SoC supports
> power floor notification. When enabled this shows the current
> status of power floor.
>
> The power floor implementation provides interfaces, which are called
> from the sysfs callbacks to enable/disable and read power floor
> status. Also provides two additional interface to check if the current
> processor thermal device interrupt is for power floor status and
> send notification to user space.
>
> Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@xxxxxxxxxxxxxxx>
> ---
> .../driver-api/thermal/intel_dptf.rst | 8 ++
> .../thermal/intel/int340x_thermal/Makefile | 1 +
> .../processor_thermal_device.c | 68 ++++++++++-
> .../processor_thermal_device.h | 8 ++
> .../processor_thermal_power_floor.c | 114 ++++++++++++++++++
> 5 files changed, 198 insertions(+), 1 deletion(-)
> create mode 100644 drivers/thermal/intel/int340x_thermal/processor_thermal_power_floor.c
>
> diff --git a/Documentation/driver-api/thermal/intel_dptf.rst b/Documentation/driver-api/thermal/intel_dptf.rst
> index 2d11e74ac665..55d90eafd94b 100644
> --- a/Documentation/driver-api/thermal/intel_dptf.rst
> +++ b/Documentation/driver-api/thermal/intel_dptf.rst
> @@ -164,6 +164,14 @@ ABI.
> ``power_limit_1_tmax_us`` (RO)
> Maximum powercap sysfs constraint_1_time_window_us for Intel RAPL
>
> +``power_floor_status`` (RO)
> + When set to 1, hardware is not able to satisfy the requested power limit
> + via Intel RAPL.
> +
> +``power_floor_enable`` (RW)
> + When set to 1, enable reading and notification of power floor status.
> + Notifications are issued for changes in the power_floor_status attribute.
> +
> :file:`/sys/bus/pci/devices/0000\:00\:04.0/`
>
> ``tcc_offset_degree_celsius`` (RW)
> diff --git a/drivers/thermal/intel/int340x_thermal/Makefile b/drivers/thermal/intel/int340x_thermal/Makefile
> index f33a3ad3bef3..fe3f43924525 100644
> --- a/drivers/thermal/intel/int340x_thermal/Makefile
> +++ b/drivers/thermal/intel/int340x_thermal/Makefile
> @@ -12,5 +12,6 @@ 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-$(CONFIG_INT340X_THERMAL) += processor_thermal_power_floor.o
> obj-$(CONFIG_INT3406_THERMAL) += int3406_thermal.o
> obj-$(CONFIG_ACPI_THERMAL_REL) += acpi_thermal_rel.o
> diff --git a/drivers/thermal/intel/int340x_thermal/processor_thermal_device.c b/drivers/thermal/intel/int340x_thermal/processor_thermal_device.c
> index 29ed7d0f7022..649f67fdf345 100644
> --- a/drivers/thermal/intel/int340x_thermal/processor_thermal_device.c
> +++ b/drivers/thermal/intel/int340x_thermal/processor_thermal_device.c
> @@ -26,6 +26,48 @@ static ssize_t power_limit_##index##_##suffix##_show(struct device *dev, \
> (unsigned long)proc_dev->power_limits[index].suffix * 1000); \
> }
>
> +static ssize_t power_floor_status_show(struct device *dev,
> + struct device_attribute *attr,
> + char *buf)
> +{
> + struct proc_thermal_device *proc_dev = dev_get_drvdata(dev);
> + int ret;
> +
> + ret = proc_thermal_read_power_floor_status(proc_dev);
> +
> + return sysfs_emit(buf, "%d\n", ret);
> +}
> +
> +static ssize_t power_floor_enable_show(struct device *dev,
> + struct device_attribute *attr,
> + char *buf)
> +{
> + struct proc_thermal_device *proc_dev = dev_get_drvdata(dev);
> + bool ret;
> +
> + ret = proc_thermal_power_floor_get_state(proc_dev);
> +
> + return sysfs_emit(buf, "%d\n", ret);
> +}
> +
> +static ssize_t power_floor_enable_store(struct device *dev,
> + struct device_attribute *attr,
> + const char *buf, size_t count)
> +{
> + struct proc_thermal_device *proc_dev = dev_get_drvdata(dev);
> + u8 state;
> + int ret;
> +
> + if (kstrtou8(buf, 0, &state))
> + return -EINVAL;
> +
> + ret = proc_thermal_power_floor_set_state(proc_dev, !!state);
> + if (ret)
> + return ret;
> +
> + return count;
> +}
> +
> POWER_LIMIT_SHOW(0, min_uw)
> POWER_LIMIT_SHOW(0, max_uw)
> POWER_LIMIT_SHOW(0, step_uw)
> @@ -50,6 +92,9 @@ static DEVICE_ATTR_RO(power_limit_1_step_uw);
> static DEVICE_ATTR_RO(power_limit_1_tmin_us);
> static DEVICE_ATTR_RO(power_limit_1_tmax_us);
>
> +static DEVICE_ATTR_RO(power_floor_status);
> +static DEVICE_ATTR_RW(power_floor_enable);
> +
> static struct attribute *power_limit_attrs[] = {
> &dev_attr_power_limit_0_min_uw.attr,
> &dev_attr_power_limit_1_min_uw.attr,
> @@ -61,12 +106,30 @@ static struct attribute *power_limit_attrs[] = {
> &dev_attr_power_limit_1_tmin_us.attr,
> &dev_attr_power_limit_0_tmax_us.attr,
> &dev_attr_power_limit_1_tmax_us.attr,
> + &dev_attr_power_floor_status.attr,
> + &dev_attr_power_floor_enable.attr,
> NULL
> };
>
> +static umode_t power_limit_attr_visible(struct kobject *kobj, struct attribute *attr, int unused)
> +{
> + struct device *dev = kobj_to_dev(kobj);
> + struct proc_thermal_device *proc_dev;
> +
> + if (attr != &dev_attr_power_floor_status.attr && attr != &dev_attr_power_floor_enable.attr)
> + return attr->mode;
> +
> + proc_dev = dev_get_drvdata(dev);
> + if (!proc_dev || !(proc_dev->mmio_feature_mask & PROC_THERMAL_FEATURE_POWER_FLOOR))
> + return 0;
> +
> + return attr->mode;
> +}
> +
> static const struct attribute_group power_limit_attribute_group = {
> .attrs = power_limit_attrs,
> - .name = "power_limits"
> + .name = "power_limits",
> + .is_visible = power_limit_attr_visible,
> };
>
> static ssize_t tcc_offset_degree_celsius_show(struct device *dev,
> @@ -380,6 +443,9 @@ 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_POWER_FLOOR)
> + proc_thermal_power_floor_set_state(proc_priv, false);
> +
> 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)
> diff --git a/drivers/thermal/intel/int340x_thermal/processor_thermal_device.h b/drivers/thermal/intel/int340x_thermal/processor_thermal_device.h
> index dd025c8c2bac..89e98f025a49 100644
> --- a/drivers/thermal/intel/int340x_thermal/processor_thermal_device.h
> +++ b/drivers/thermal/intel/int340x_thermal/processor_thermal_device.h
> @@ -63,6 +63,7 @@ struct rapl_mmio_regs {
> #define PROC_THERMAL_FEATURE_WT_REQ 0x08
> #define PROC_THERMAL_FEATURE_DLVR 0x10
> #define PROC_THERMAL_FEATURE_WT_HINT 0x20
> +#define PROC_THERMAL_FEATURE_POWER_FLOOR 0x40
>
> #if IS_ENABLED(CONFIG_PROC_THERMAL_MMIO_RAPL)
> int proc_thermal_rapl_add(struct pci_dev *pdev, struct proc_thermal_device *proc_priv);
> @@ -91,6 +92,13 @@ void proc_thermal_wt_req_remove(struct pci_dev *pdev);
> #define MBOX_DATA_BIT_AC_DC 30
> #define MBOX_DATA_BIT_VALID 31
>
> +int proc_thermal_read_power_floor_status(struct proc_thermal_device *proc_priv);
> +int proc_thermal_power_floor_set_state(struct proc_thermal_device *proc_priv, bool enable);
> +bool proc_thermal_power_floor_get_state(struct proc_thermal_device *proc_priv);
> +void proc_thermal_power_floor_intr_callback(struct pci_dev *pdev,
> + struct proc_thermal_device *proc_priv);
> +bool proc_thermal_check_power_floor_intr(struct proc_thermal_device *proc_priv);
> +
> 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,
> diff --git a/drivers/thermal/intel/int340x_thermal/processor_thermal_power_floor.c b/drivers/thermal/intel/int340x_thermal/processor_thermal_power_floor.c
> new file mode 100644
> index 000000000000..cac3a38f430f
> --- /dev/null
> +++ b/drivers/thermal/intel/int340x_thermal/processor_thermal_power_floor.c
> @@ -0,0 +1,114 @@
> +// SPDX-License-Identifier: GPL-2.0-only
> +/*
> + * Processor thermal device module for registering and processing
> + * power floor. When the hardware reduces the power to the minimum
> + * possible, the power floor is notified via an interrupt.
> + *
> + * Operation:
> + * When user space enables power floor reporting:
> + * - Use mailbox to:
> + * Enable processor thermal device interrupt
> + *
> + * - Current status of power floor is read from offset 0x5B18
> + * bit 39.
> + *
> + * Two interface functions are provided to call when there is a
> + * thermal device interrupt:
> + * - proc_thermal_power_floor_intr():
> + * Check if the interrupt is for change in power floor.
> + * Called from interrupt context.
> + *
> + * - proc_thermal_power_floor_intr_callback():
> + * Callback for interrupt processing in thread context. This involves
> + * sending notification to user space that there is a change in the
> + * power floor status.
> + *
> + * Copyright (c) 2023, Intel Corporation.
> + */
> +
> +#include <linux/pci.h>
> +#include "processor_thermal_device.h"
> +
> +#define SOC_POWER_FLOOR_INT_STATUS_OFF 0x5B18

"OFFSET" would be better than "OFF" IMO.

> +#define SOC_POWER_FLOOR_STATUS BIT(39)
> +#define SOC_POWER_FLOOR_SHIFT 39
> +
> +#define SOC_POWER_FLOOR_INT_ENABLE_BIT 31
> +#define SOC_POWER_FLOOR_INT_ACTIVE BIT(3)
> +
> +/* Mark time windows as invalid as this is not applicable */

The meaning of the above comment is unclear.

> +#define SOC_POWER_FLOOR_TIME_WINDOW -1
> +
> +int proc_thermal_read_power_floor_status(struct proc_thermal_device *proc_priv)
> +{
> + u64 status = 0;
> +
> + status = readq(proc_priv->mmio_base + SOC_POWER_FLOOR_INT_STATUS_OFF);
> + return (status & SOC_POWER_FLOOR_STATUS) >> SOC_POWER_FLOOR_SHIFT;
> +}
> +EXPORT_SYMBOL_NS_GPL(proc_thermal_read_power_floor_status, INT340X_THERMAL);
> +
> +static bool enable_state;
> +static DEFINE_MUTEX(pf_lock);
> +
> +int proc_thermal_power_floor_set_state(struct proc_thermal_device *proc_priv, bool enable)
> +{
> + int ret = 0;
> +
> + mutex_lock(&pf_lock);
> + if (enable_state == enable)
> + goto pf_unlock;
> +
> + ret = processor_thermal_mbox_interrupt_config(to_pci_dev(proc_priv->dev), enable,
> + SOC_POWER_FLOOR_INT_ENABLE_BIT,
> + SOC_POWER_FLOOR_TIME_WINDOW);
> + if (!ret)
> + enable_state = enable;
> +
> +pf_unlock:
> + mutex_unlock(&pf_lock);
> +
> + return ret;
> +}
> +EXPORT_SYMBOL_NS_GPL(proc_thermal_power_floor_set_state, INT340X_THERMAL);
> +
> +bool proc_thermal_power_floor_get_state(struct proc_thermal_device *proc_priv)
> +{
> + return enable_state;
> +}
> +EXPORT_SYMBOL_NS_GPL(proc_thermal_power_floor_get_state, INT340X_THERMAL);
> +
> +/*
> + * Callback to check if interrupt for power floor is active.
> + * Caution: Called from interrupt context.
> + */

I would prefer this comment and the one below to be proper kerneldoc
ones or to be moved inside the functions.

> +bool proc_thermal_check_power_floor_intr(struct proc_thermal_device *proc_priv)
> +{
> + u64 int_status;
> +
> + int_status = readq(proc_priv->mmio_base + SOC_POWER_FLOOR_INT_STATUS_OFF);
> + if (int_status & SOC_POWER_FLOOR_INT_ACTIVE)
> + return true;
> +
> + return false;

return !!(int_status & SOC_POWER_FLOOR_INT_ACTIVE);

> +}
> +EXPORT_SYMBOL_NS_GPL(proc_thermal_check_power_floor_intr, INT340X_THERMAL);
> +
> +/* Callback to notify user space */

Notify user space about what and how? As is, this comment isn't very
useful IMO.

> +void proc_thermal_power_floor_intr_callback(struct pci_dev *pdev,
> + struct proc_thermal_device *proc_priv)
> +{
> + u64 status;
> +
> + status = readq(proc_priv->mmio_base + SOC_POWER_FLOOR_INT_STATUS_OFF);
> + if (!(status & SOC_POWER_FLOOR_INT_ACTIVE))
> + return;
> +
> + writeq(status & ~SOC_POWER_FLOOR_INT_ACTIVE,
> + proc_priv->mmio_base + SOC_POWER_FLOOR_INT_STATUS_OFF);
> + sysfs_notify(&pdev->dev.kobj, "power_limits", "power_floor_status");
> +}
> +EXPORT_SYMBOL_NS_GPL(proc_thermal_power_floor_intr_callback, INT340X_THERMAL);
> +
> +MODULE_IMPORT_NS(INT340X_THERMAL);
> +MODULE_LICENSE("GPL");
> --