[PATCH v3 5/6] drm/amd/pm: Add temperature metrics sysfs entry

[Lazar, Lijo]

On 8/4/2025 7:47 PM, Asad Kamal wrote:
> Add temperature metrics sysfs entry to expose gpuboard/baseboard
> temperature metrics
> 
> v2: Removed unused function, rename functions(Lijo)
> 
> v3: Remove unnecessary initialization
> 
> Signed-off-by: Asad Kamal <asad.kamal at amd.com>
> ---
>  drivers/gpu/drm/amd/pm/amdgpu_pm.c | 135 +++++++++++++++++++++++++++++
>  1 file changed, 135 insertions(+)
> 
> diff --git a/drivers/gpu/drm/amd/pm/amdgpu_pm.c b/drivers/gpu/drm/amd/pm/amdgpu_pm.c
> index 8d934a365c3b..b94cdfd9e6ae 100644
> --- a/drivers/gpu/drm/amd/pm/amdgpu_pm.c
> +++ b/drivers/gpu/drm/amd/pm/amdgpu_pm.c
> @@ -2080,6 +2080,134 @@ static int pp_dpm_clk_default_attr_update(struct amdgpu_device *adev, struct amd
>  	return 0;
>  }
>  
> +/**
> + * DOC: board
> + *
> + * Certain SOCs can support various board attributes reporting. This is useful
> + * for user application to monitor various bard reated attributes.

Typo bard -> board.

> + *
> + * The amdgpu driver provides a sysfs API for reporting board attributes. Presently,
> + * only two types of attributes are reported, baseboard temperature and
> + * gpu board temperature. Both of them are reported as binary files.
> + *
> + * * .. code-block:: console
> + *
> + *      hexdump /sys/bus/pci/devices/.../board/baseboard_temp
> + *
> + *      hexdump /sys/bus/pci/devices/.../board/gpuboard_temp
> + *
> + */
> +
> +/**
> + * DOC: baseboard_temp
> + *
> + * The amdgpu driver provides a sysfs API for retrieving current baseboard
> + * temperature metrics data. The file baseboard_temp is used for this.
> + * Reading the file will dump all the current baseboard tempertature  metrics data.

Typo -> temperature
> + */
> +static ssize_t amdgpu_get_baseboard_temp_metrics(struct device *dev,
> +						 struct device_attribute *attr, char *buf)
> +{
> +	struct drm_device *ddev = dev_get_drvdata(dev);
> +	struct amdgpu_device *adev = drm_to_adev(ddev);
> +	ssize_t size;
> +	int ret;
> +
> +	ret = amdgpu_pm_get_access_if_active(adev);
> +	if (ret)
> +		return ret;
> +
> +	size = amdgpu_dpm_get_temp_metrics(adev, SMU_TEMP_METRIC_BASEBOARD, NULL);
> +	if (size <= 0)
> +		goto out;
> +	if (size >= PAGE_SIZE) {
> +		ret = -ENOSPC;
> +		goto out;
> +	}
> +
> +	amdgpu_dpm_get_temp_metrics(adev, SMU_TEMP_METRIC_BASEBOARD, buf);
> +
> +out:
> +	amdgpu_pm_put_access(adev);
> +
> +	if (ret)
> +		return ret;
> +
> +	return size;
> +}
> +
> +/**
> + * DOC: gpuboard_temp
> + *
> + * The amdgpu driver provides a sysfs API for retrieving current gpuboard
> + * temperature metrics data. The file gpuboard_temp is used for this.
> + * Reading the file will dump all the current gpuboard tempertature  metrics data.


Typo -> temperature

Thanks,
Lijo

> + */
> +static ssize_t amdgpu_get_gpuboard_temp_metrics(struct device *dev,
> +						struct device_attribute *attr, char *buf)
> +{
> +	struct drm_device *ddev = dev_get_drvdata(dev);
> +	struct amdgpu_device *adev = drm_to_adev(ddev);
> +	ssize_t size;
> +	int ret;
> +
> +	ret = amdgpu_pm_get_access_if_active(adev);
> +	if (ret)
> +		return ret;
> +
> +	size = amdgpu_dpm_get_temp_metrics(adev, SMU_TEMP_METRIC_GPUBOARD, NULL);
> +	if (size <= 0)
> +		goto out;
> +	if (size >= PAGE_SIZE) {
> +		ret = -ENOSPC;
> +		goto out;
> +	}
> +
> +	amdgpu_dpm_get_temp_metrics(adev, SMU_TEMP_METRIC_GPUBOARD, buf);
> +
> +out:
> +	amdgpu_pm_put_access(adev);
> +
> +	if (ret)
> +		return ret;
> +
> +	return size;
> +}
> +
> +static DEVICE_ATTR(baseboard_temp, 0444, amdgpu_get_baseboard_temp_metrics, NULL);
> +static DEVICE_ATTR(gpuboard_temp, 0444, amdgpu_get_gpuboard_temp_metrics, NULL);
> +
> +static struct attribute *board_attrs[] = {
> +	&dev_attr_baseboard_temp.attr,
> +	&dev_attr_gpuboard_temp.attr,
> +	NULL
> +};
> +
> +static umode_t amdgpu_board_attr_visible(struct kobject *kobj, struct attribute *attr, int n)
> +{
> +	struct device *dev = kobj_to_dev(kobj);
> +	struct drm_device *ddev = dev_get_drvdata(dev);
> +	struct amdgpu_device *adev = drm_to_adev(ddev);
> +
> +	if (attr == &dev_attr_baseboard_temp.attr) {
> +		if (!amdgpu_dpm_is_temp_metrics_supported(adev, SMU_TEMP_METRIC_BASEBOARD))
> +			return 0;
> +	}
> +
> +	if (attr == &dev_attr_gpuboard_temp.attr) {
> +		if (!amdgpu_dpm_is_temp_metrics_supported(adev, SMU_TEMP_METRIC_GPUBOARD))
> +			return 0;
> +	}
> +
> +	return attr->mode;
> +}
> +
> +const struct attribute_group amdgpu_board_attr_group = {
> +	.name = "board",
> +	.attrs = board_attrs,
> +	.is_visible = amdgpu_board_attr_visible,
> +};
> +
>  /* pm policy attributes */
>  struct amdgpu_pm_policy_attr {
>  	struct device_attribute dev_attr;
> @@ -4468,6 +4596,13 @@ int amdgpu_pm_sysfs_init(struct amdgpu_device *adev)
>  			goto err_out0;
>  	}
>  
> +	if (amdgpu_dpm_is_temp_metrics_supported(adev, SMU_TEMP_METRIC_GPUBOARD)) {
> +		ret = devm_device_add_group(adev->dev,
> +					    &amdgpu_board_attr_group);
> +		if (ret)
> +			goto err_out0;
> +	}
> +
>  	adev->pm.sysfs_initialized = true;
>  
>  	return 0;