Re: [PATCH v1 04/15] platform/x86/amd/pmf: Add support SPS PMF feature

[Hans de Goede <hdegoede@redhat.com>]

Hi,

On 7/12/22 16:58, Shyam Sundar S K wrote:
> SPS (a.k.a. Static Power Slider) gives a feel of Windows performance
> power slider for the Linux users, where the user selects a certain
> mode (like "balanced", "low-power" or "performance") and the thermals
> associated with each selected mode gets applied from the silicon
> side via the mailboxes defined through PMFW.
> 
> PMF driver hooks to platform_profile by reading the PMF ACPI fn9 to
> see if the support is being advertised by ACPI interface.
> 
> If supported, the PMF driver reacts to platform_profile selection choices
> made by the user and adjust the system thermal behavior.
> 
> Signed-off-by: Shyam Sundar S K <Shyam-sundar.S-k@amd.com>
> ---
>  drivers/platform/x86/amd/pmf/Makefile |   2 +-
>  drivers/platform/x86/amd/pmf/acpi.c   |  29 +++++
>  drivers/platform/x86/amd/pmf/core.c   |  26 ++++
>  drivers/platform/x86/amd/pmf/pmf.h    |  65 ++++++++++
>  drivers/platform/x86/amd/pmf/sps.c    | 166 ++++++++++++++++++++++++++
>  5 files changed, 287 insertions(+), 1 deletion(-)
>  create mode 100644 drivers/platform/x86/amd/pmf/sps.c
> 
> diff --git a/drivers/platform/x86/amd/pmf/Makefile b/drivers/platform/x86/amd/pmf/Makefile
> index 2617eba773ce..557521a80427 100644
> --- a/drivers/platform/x86/amd/pmf/Makefile
> +++ b/drivers/platform/x86/amd/pmf/Makefile
> @@ -5,4 +5,4 @@
>  #
>  
>  obj-$(CONFIG_AMD_PMF) += amd-pmf.o
> -amd-pmf-objs := core.o acpi.o
> +amd-pmf-objs := core.o acpi.o sps.o
> diff --git a/drivers/platform/x86/amd/pmf/acpi.c b/drivers/platform/x86/amd/pmf/acpi.c
> index addcaae5675c..fd5ab7a116f0 100644
> --- a/drivers/platform/x86/amd/pmf/acpi.c
> +++ b/drivers/platform/x86/amd/pmf/acpi.c
> @@ -54,6 +54,7 @@ static union acpi_object *apmf_if_call(struct apmf_if *apmf_if, int func, struct
>  static void apmf_if_parse_functions(struct apmf_if_functions *func, u32 mask)
>  {
>  	func->system_params = mask & APMF_FUNC_GET_SYS_PARAMS;
> +	func->static_slider_granular = mask & APMF_FUNC_STATIC_SLIDER_GRANULAR;
>  }
>  
>  static int apmf_if_verify_interface(struct amd_pmf_dev *pdev, struct apmf_if *ampf_if)
> @@ -130,6 +131,34 @@ static int apmf_get_system_params(struct apmf_if *ampf_if)
>  	return err;
>  }
>  
> +int apmf_get_static_slider_granular(struct apmf_if *ampf_if,
> +				    struct apmf_static_slider_granular_output *data)
> +{
> +	union acpi_object *info;
> +	size_t size;
> +	int err = 0;
> +
> +	info = apmf_if_call(ampf_if, APMF_FUNC_STATIC_SLIDER_GRANULAR, NULL);
> +	if (!info)
> +		return -EIO;
> +
> +	size = *(u16 *)info->buffer.pointer;
> +
> +	if (size < sizeof(*data)) {
> +		pr_err("PMF: buffer too small %zu\n", size);
> +		err = -EINVAL;
> +		goto out;
> +	}
> +
> +	size = min(sizeof(*data), size);
> +	memset(data, 0, sizeof(*data));
> +	memcpy(data, info->buffer.pointer, size);

Same remarks wrt size handling as in patch 3/15. As mentioned you really need
a helper for this. Something with a prototype like this:

int apmf_if_call_store_buffer(struct apmf_if *ampf_if, int func, void *dest, size_t dest_size);

And then this entire function becomes just:

int apmf_get_static_slider_granular(struct apmf_if *ampf_if,
				    struct apmf_static_slider_granular_output *data)
	return apmf_if_call_store_buffer(ampf_if, APMF_FUNC_STATIC_SLIDER_GRANULAR, data, sizeof(*data));
}


> +
> +out:
> +	kfree(info);
> +	return err;
> +}
> +
>  static acpi_handle apmf_if_probe_handle(void)
>  {
>  	acpi_handle handle = NULL;
> diff --git a/drivers/platform/x86/amd/pmf/core.c b/drivers/platform/x86/amd/pmf/core.c
> index c5002b7bb904..b6006e8ee1a1 100644
> --- a/drivers/platform/x86/amd/pmf/core.c
> +++ b/drivers/platform/x86/amd/pmf/core.c
> @@ -12,6 +12,7 @@
>  #include <linux/module.h>
>  #include <linux/pci.h>
>  #include <linux/platform_device.h>
> +#include <linux/power_supply.h>
>  #include "pmf.h"
>  
>  /* PMF-SMU communication registers */
> @@ -45,6 +46,14 @@
>  #define DELAY_MIN_US	2000
>  #define DELAY_MAX_US	3000
>  
> +int amd_pmf_get_power_source(void)
> +{
> +	if (power_supply_is_system_supplied() > 0)
> +		return POWER_SOURCE_AC;
> +	else
> +		return POWER_SOURCE_DC;
> +}
> +>  static inline u32 amd_pmf_reg_read(struct amd_pmf_dev *dev, int reg_offset)
>  {
>  	return ioread32(dev->regbase + reg_offset);
> @@ -138,6 +147,21 @@ static const struct pci_device_id pmf_pci_ids[] = {
>  	{ }
>  };
>  
> +static void amd_pmf_init_features(struct amd_pmf_dev *dev)
> +{
> +	/* Enable Static Slider */
> +	if (is_apmf_func_supported(APMF_FUNC_STATIC_SLIDER_GRANULAR)) {

Just put a:

	if (!is_apmf_func_supported(APMF_FUNC_STATIC_SLIDER_GRANULAR))
		return;

check at the top of amd_pmf_init_sps(dev); and then you can drop this
entire function.

> +		amd_pmf_init_sps(dev);
> +		dev_dbg(dev->dev, "SPS enabled and Platform Profiles registered\n");
> +	}
> +}
> +
> +static void amd_pmf_deinit_features(struct amd_pmf_dev *dev)
> +{
> +	if (is_apmf_func_supported(APMF_FUNC_STATIC_SLIDER_GRANULAR))
> +		amd_pmf_deinit_sps(dev);

Same here.

> +}
> +
>  static const struct acpi_device_id amd_pmf_acpi_ids[] = {
>  	{"AMDI0102", 0},
>  	{ }
> @@ -206,6 +230,7 @@ static int amd_pmf_probe(struct platform_device *pdev)
>  
>  	apmf_acpi_init(dev);
>  	platform_set_drvdata(pdev, dev);
> +	amd_pmf_init_features(dev);
>  
>  	mutex_init(&dev->lock);
>  	dev_info(dev->dev, "registered PMF device successfully\n");
> @@ -218,6 +243,7 @@ static int amd_pmf_remove(struct platform_device *pdev)
>  	struct amd_pmf_dev *dev = platform_get_drvdata(pdev);
>  
>  	mutex_destroy(&dev->lock);
> +	amd_pmf_deinit_features(dev);
>  	kfree(dev->buf);
>  	return 0;
>  }
> diff --git a/drivers/platform/x86/amd/pmf/pmf.h b/drivers/platform/x86/amd/pmf/pmf.h
> index 40f038eb6197..a405987ae653 100644
> --- a/drivers/platform/x86/amd/pmf/pmf.h
> +++ b/drivers/platform/x86/amd/pmf/pmf.h
> @@ -7,14 +7,17 @@
>   *
>   * Author: Shyam Sundar S K <Shyam-sundar.S-k@amd.com>
>   */
> +
>  #ifndef PMF_H
>  #define PMF_H
>  
>  #include <linux/acpi.h>
> +#include <linux/platform_profile.h>
>  
>  /* APMF Functions */
>  #define APMF_FUNC_VERIFY_INTERFACE			0
>  #define APMF_FUNC_GET_SYS_PARAMS			1
> +#define APMF_FUNC_STATIC_SLIDER_GRANULAR	9
>  
>  /* Message Definitions */
>  #define SET_SPL				0x03 /* SPL: Sustained Power Limit */
> @@ -35,9 +38,12 @@
>  #define GET_STT_LIMIT_APU	0x20
>  #define GET_STT_LIMIT_HS2	0x21
>  
> +#define ARG_NONE 0
> +
>  /* AMD PMF BIOS interfaces */
>  struct apmf_if_functions {
>  	bool system_params;
> +	bool static_slider_granular;
>  };
>  
>  struct apmf_if {
> @@ -59,6 +65,30 @@ struct apmf_system_params {
>  	u8 command_code;
>  } __packed;
>  
> +enum amd_stt_skin_temp {
> +	STT_TEMP_APU,
> +	STT_TEMP_HS2,
> +	STT_TEMP_COUNT,
> +};
> +
> +enum amd_slider_op {
> +	SLIDER_OP_GET,
> +	SLIDER_OP_SET,
> +};
> +
> +enum power_source {
> +	POWER_SOURCE_AC,
> +	POWER_SOURCE_DC,
> +	POWER_SOURCE_MAX,
> +};
> +
> +enum power_modes {
> +	POWER_MODE_PERFORMANCE,
> +	POWER_MODE_BALANCED_POWER,
> +	POWER_MODE_POWER_SAVER,
> +	POWER_MODE_MAX,
> +};
> +
>  struct amd_pmf_dev {
>  	void __iomem *regbase;
>  	void __iomem *smu_virt_addr;
> @@ -69,11 +99,46 @@ struct amd_pmf_dev {
>  	u32 low;
>  	struct device *dev;
>  	struct apmf_if *apmf_if;
> +	enum platform_profile_option current_profile;
> +	struct platform_profile_handler pprof;
>  	struct mutex lock; /* protects the PMF interface */
>  };
>  
> +struct apmf_sps_prop_granular {
> +	u32 fppt;
> +	u32 sppt;
> +	u32 sppt_apu_only;
> +	u32 spl;
> +	u32 stt_min;
> +	u8 stt_skin_temp[STT_TEMP_COUNT];
> +	u32 fan_id;
> +} __packed;
> +
> +/* Static Slider */
> +struct apmf_static_slider_granular_output {
> +	u16 size;
> +	struct apmf_sps_prop_granular prop[POWER_SOURCE_MAX * POWER_MODE_MAX];
> +};

missing __packed here I presume, since this is the data returned
by an ACPI call ?

> +
> +struct amd_pmf_static_slider_granular {
> +	u16 size;
> +	struct apmf_sps_prop_granular prop[POWER_SOURCE_MAX][POWER_MODE_MAX];
> +};
> +
>  /* Core Layer */
>  int apmf_acpi_init(struct amd_pmf_dev *pmf_dev);
> +int is_apmf_func_supported(unsigned long index);
>  int amd_pmf_send_cmd(struct amd_pmf_dev *dev, u8 message, bool get, u32 arg, u32 *data);
> +int amd_pmf_get_power_source(void);
> +
> +/* SPS Layer */
> +u8 amd_pmf_get_pprof_modes(struct amd_pmf_dev *pmf);
> +void amd_pmf_update_slider(struct amd_pmf_dev *dev, bool op, int idx,
> +			   struct amd_pmf_static_slider_granular *table);
> +int amd_pmf_init_sps(struct amd_pmf_dev *dev);
> +void amd_pmf_deinit_sps(struct amd_pmf_dev *dev);
> +int apmf_get_static_slider_granular(struct apmf_if *ampf_if,
> +				    struct apmf_static_slider_granular_output *output);
> +void amd_pmf_load_defaults_sps(struct amd_pmf_dev *dev);
>  
>  #endif /* PMF_H */
> diff --git a/drivers/platform/x86/amd/pmf/sps.c b/drivers/platform/x86/amd/pmf/sps.c
> new file mode 100644
> index 000000000000..25289cf5608c
> --- /dev/null
> +++ b/drivers/platform/x86/amd/pmf/sps.c
> @@ -0,0 +1,166 @@
> +// SPDX-License-Identifier: GPL-2.0
> +/*
> + * AMD Platform Management Framework (PMF) Driver
> + *
> + * Copyright (c) 2022, Advanced Micro Devices, Inc.
> + * All Rights Reserved.
> + *
> + * Author: Shyam Sundar S K <Shyam-sundar.S-k@amd.com>
> + */
> +
> +#include "pmf.h"
> +
> +static int amd_pmf_profile_get(struct platform_profile_handler *pprof,
> +			       enum platform_profile_option *profile);
> +static int amd_pmf_profile_set(struct platform_profile_handler *pprof,
> +			       enum platform_profile_option profile);
> +static struct amd_pmf_static_slider_granular config_store;
> +
> +void amd_pmf_load_defaults_sps(struct amd_pmf_dev *dev)
> +{
> +	struct apmf_static_slider_granular_output output;
> +	int i, j, idx = 0;
> +
> +	memset(&config_store, 0, sizeof(config_store));
> +
> +	if (dev->apmf_if->func.static_slider_granular) {
> +		apmf_get_static_slider_granular(dev->apmf_if, &output);
> +
> +		for (i = 0; i < POWER_SOURCE_MAX; i++) {
> +			for (j = 0; j < POWER_MODE_MAX; j++) {
> +				config_store.prop[i][j].spl = output.prop[idx].spl;
> +				config_store.prop[i][j].sppt = output.prop[idx].sppt;
> +				config_store.prop[i][j].sppt_apu_only =
> +							output.prop[idx].sppt_apu_only;
> +				config_store.prop[i][j].fppt = output.prop[idx].fppt;
> +				config_store.prop[i][j].stt_min = output.prop[idx].stt_min;
> +				config_store.prop[i][j].stt_skin_temp[STT_TEMP_APU] =
> +						output.prop[idx].stt_skin_temp[STT_TEMP_APU];
> +				config_store.prop[i][j].stt_skin_temp[STT_TEMP_HS2] =
> +						output.prop[idx].stt_skin_temp[STT_TEMP_HS2];
> +				config_store.prop[i][j].fan_id = output.prop[idx].fan_id;
> +				idx++;
> +			}
> +		}
> +	}
> +}
> +
> +void amd_pmf_update_slider(struct amd_pmf_dev *dev, bool op, int idx,
> +			   struct amd_pmf_static_slider_granular *table)
> +{
> +	int src = amd_pmf_get_power_source();
> +
> +	if (op == SLIDER_OP_SET) {
> +		amd_pmf_send_cmd(dev, SET_SPL, false, config_store.prop[src][idx].spl, NULL);
> +		amd_pmf_send_cmd(dev, SET_FPPT, false, config_store.prop[src][idx].fppt, NULL);
> +		amd_pmf_send_cmd(dev, SET_SPPT, false, config_store.prop[src][idx].sppt, NULL);
> +		amd_pmf_send_cmd(dev, SET_SPPT_APU_ONLY, false,
> +				 config_store.prop[src][idx].sppt_apu_only, NULL);
> +		amd_pmf_send_cmd(dev, SET_STT_MIN_LIMIT, false,
> +				 config_store.prop[src][idx].stt_min, NULL);
> +		amd_pmf_send_cmd(dev, SET_STT_LIMIT_APU, false,
> +				 config_store.prop[src][idx].stt_skin_temp[STT_TEMP_APU], NULL);
> +		amd_pmf_send_cmd(dev, SET_STT_LIMIT_HS2, false,
> +				 config_store.prop[src][idx].stt_skin_temp[STT_TEMP_HS2], NULL);
> +	} else if (op == SLIDER_OP_GET) {
> +		amd_pmf_send_cmd(dev, GET_SPL, true, ARG_NONE, &table->prop[src][idx].spl);
> +		amd_pmf_send_cmd(dev, GET_FPPT, true, ARG_NONE, &table->prop[src][idx].fppt);
> +		amd_pmf_send_cmd(dev, GET_SPPT, true, ARG_NONE, &table->prop[src][idx].sppt);
> +		amd_pmf_send_cmd(dev, GET_SPPT_APU_ONLY, true, ARG_NONE,
> +				 &table->prop[src][idx].sppt_apu_only);
> +		amd_pmf_send_cmd(dev, GET_STT_MIN_LIMIT, true, ARG_NONE,
> +				 &table->prop[src][idx].stt_min);
> +		amd_pmf_send_cmd(dev, GET_STT_LIMIT_APU, true, ARG_NONE,
> +				 (u32 *)&table->prop[src][idx].stt_skin_temp[STT_TEMP_APU]);
> +		amd_pmf_send_cmd(dev, GET_STT_LIMIT_HS2, true, ARG_NONE,
> +				 (u32 *)&table->prop[src][idx].stt_skin_temp[STT_TEMP_HS2]);
> +	}
> +}
> +
> +static int amd_pmf_profile_get(struct platform_profile_handler *pprof,
> +			       enum platform_profile_option *profile)
> +{
> +	struct amd_pmf_dev *pmf = container_of(pprof, struct amd_pmf_dev, pprof);
> +
> +	*profile = pmf->current_profile;
> +	return 0;
> +}
> +
> +u8 amd_pmf_get_pprof_modes(struct amd_pmf_dev *pmf)
> +{
> +	u8 mode;
> +
> +	switch (pmf->current_profile) {
> +	case PLATFORM_PROFILE_PERFORMANCE:
> +		mode = 0;

Please use POWER_MODE_PERFORMANCE instead of 1 here.

> +		break;
> +	case PLATFORM_PROFILE_BALANCED:
> +		mode = 1;

POWER_MODE_BALANCED_POWER instead of 1.

> +		break;
> +	case PLATFORM_PROFILE_LOW_POWER:
> +		mode = 2;

POWER_MODE_POWER_SAVER instead of 2.

> +		break;
> +	default:
> +		dev_err(pmf->dev, "Unknown Platform Profile.\n");
> +		break;
> +	}
> +
> +	return mode;
> +}
> +
> +int amd_pmf_profile_set(struct platform_profile_handler *pprof,
> +			enum platform_profile_option profile)
> +{
> +	struct amd_pmf_dev *pmf = container_of(pprof, struct amd_pmf_dev, pprof);
> +	u8 mode;
> +
> +	pmf->current_profile = profile;
> +	mode = amd_pmf_get_pprof_modes(pmf);
> +	amd_pmf_update_slider(pmf, SLIDER_OP_SET, mode, NULL);
> +	return 0;
> +}
> +
> +static void amd_pmf_profile_refresh(struct amd_pmf_dev *dev)
> +{
> +	enum platform_profile_option profile;
> +
> +	/* Set intially to Balanced Mode */
> +	profile = PLATFORM_PROFILE_BALANCED;
> +
> +	if (profile != dev->current_profile) {
> +		dev->current_profile = profile;
> +		platform_profile_notify();
> +	}
> +}
> +
> +int amd_pmf_init_sps(struct amd_pmf_dev *dev)
> +{
> +	int err = 0;
> +
> +	dev->pprof.profile_get = amd_pmf_profile_get;
> +	dev->pprof.profile_set = amd_pmf_profile_set;
> +
> +	/* Setup supported modes */
> +	set_bit(PLATFORM_PROFILE_LOW_POWER, dev->pprof.choices);
> +	set_bit(PLATFORM_PROFILE_BALANCED, dev->pprof.choices);
> +	set_bit(PLATFORM_PROFILE_PERFORMANCE, dev->pprof.choices);
> +
> +	/* Create platform_profile structure and register */
> +	err = platform_profile_register(&dev->pprof);
> +	if (err) {
> +		dev_err(dev->dev, "Failed to register SPS support, this is most likely an SBIOS bug: %d\n",
> +			err);
> +		return -EEXIST;
> +	}
> +	/* Ensure initial values are correct */
> +	amd_pmf_profile_refresh(dev);

Since dev->current_profile == 0 and thus  dev->current_profile != PLATFORM_PROFILE_BALANCED
this ends up doing 2 things:

1.	dev->current_profile = PLATFORM_PROFILE_BALANCED;
2.	platform_profile_notify()

Where 2 is not useful since the /sys/firmware/acpi/power_profile
attribute does not exist before the platform_profile_register() call
we have just done, so nothing is listening on it.

So this entire function call can be replaced with just:

	dev->current_profile = PLATFORM_PROFILE_BALANCED;

and the badly named refresh (it is not a refresh since it does
not fetch values from the hw) function can be dropped.

> +
> +	amd_pmf_load_defaults_sps(dev);
> +
> +	return err;
> +}
> +
> +void amd_pmf_deinit_sps(struct amd_pmf_dev *dev)
> +{
> +	platform_profile_remove();
> +}

Regards,

Hans