[PATCH v2] platform/x86: int3472: Evaluate device's _DSM method to control imaging clock

[PATCH v2] platform/x86: int3472: Evaluate device's _DSM method to control imaging clock

[Hans de Goede]

From: Bingbu Cao <bingbu.cao@intel.com>

On some platforms, the imaging clock should be controlled by evaluating
specific clock device's _DSM method instead of setting gpio, so this
change register clock if no gpio based clock and then use the _DSM method
to enable and disable clock.

Signed-off-by: Bingbu Cao <bingbu.cao@intel.com>
Signed-off-by: Hao Yao <hao.yao@intel.com>
Link: https://lore.kernel.org/r/20230524035135.90315-2-bingbu.cao@intel.com
---
Changes in v2 (Hans de Goede):
- Minor comment / code changes (address Andy's review remarks)
- Add a acpi_check_dsm() call
- Return 0 instead of error if we already have a GPIO clk or if
  acpi_check_dsm() fails
- Rename skl_int3472_register_clock() -> skl_int3472_register_gpio_clock()
  and name the new function: skl_int3472_register_dsm_clock()
- Move the skl_int3472_register_dsm_clock() call to after
  acpi_dev_free_resource_list() and add error checking for it
---
 .../x86/intel/int3472/clk_and_regulator.c     | 89 ++++++++++++++++++-
 drivers/platform/x86/intel/int3472/common.h   | 10 ++-
 drivers/platform/x86/intel/int3472/discrete.c |  8 +-
 3 files changed, 99 insertions(+), 8 deletions(-)

diff --git a/drivers/platform/x86/intel/int3472/clk_and_regulator.c b/drivers/platform/x86/intel/int3472/clk_and_regulator.c
index 1086c3d83494..9bcf8c64b8e7 100644
--- a/drivers/platform/x86/intel/int3472/clk_and_regulator.c
+++ b/drivers/platform/x86/intel/int3472/clk_and_regulator.c
@@ -11,6 +11,37 @@
 
 #include "common.h"
 
+/*
+ * 82c0d13a-78c5-4244-9bb1-eb8b539a8d11
+ * This _DSM GUID allows controlling the sensor clk when it is not controlled
+ * through a GPIO.
+ */
+static const guid_t img_clk_guid =
+	GUID_INIT(0x82c0d13a, 0x78c5, 0x4244,
+		  0x9b, 0xb1, 0xeb, 0x8b, 0x53, 0x9a, 0x8d, 0x11);
+
+static void skl_int3472_enable_clk_acpi_method(struct int3472_gpio_clock *clk,
+					       bool enable)
+{
+	struct int3472_discrete_device *int3472 = to_int3472_device(clk);
+	union acpi_object args[3];
+	union acpi_object argv4;
+
+	args[0].integer.type = ACPI_TYPE_INTEGER;
+	args[0].integer.value = clk->imgclk_index;
+	args[1].integer.type = ACPI_TYPE_INTEGER;
+	args[1].integer.value = enable ? 1 : 0;
+	args[2].integer.type = ACPI_TYPE_INTEGER;
+	args[2].integer.value = 1;
+
+	argv4.type = ACPI_TYPE_PACKAGE;
+	argv4.package.count = 3;
+	argv4.package.elements = args;
+
+	acpi_evaluate_dsm(acpi_device_handle(int3472->adev), &img_clk_guid,
+			  0, 1, &argv4);
+}
+
 /*
  * The regulators have to have .ops to be valid, but the only ops we actually
  * support are .enable and .disable which are handled via .ena_gpiod. Pass an
@@ -22,7 +53,11 @@ static int skl_int3472_clk_prepare(struct clk_hw *hw)
 {
 	struct int3472_gpio_clock *clk = to_int3472_clk(hw);
 
-	gpiod_set_value_cansleep(clk->ena_gpio, 1);
+	if (clk->ena_gpio)
+		gpiod_set_value_cansleep(clk->ena_gpio, 1);
+	else
+		skl_int3472_enable_clk_acpi_method(clk, 1);
+
 	return 0;
 }
 
@@ -30,7 +65,10 @@ static void skl_int3472_clk_unprepare(struct clk_hw *hw)
 {
 	struct int3472_gpio_clock *clk = to_int3472_clk(hw);
 
-	gpiod_set_value_cansleep(clk->ena_gpio, 0);
+	if (clk->ena_gpio)
+		gpiod_set_value_cansleep(clk->ena_gpio, 0);
+	else
+		skl_int3472_enable_clk_acpi_method(clk, 0);
 }
 
 static int skl_int3472_clk_enable(struct clk_hw *hw)
@@ -86,8 +124,51 @@ static const struct clk_ops skl_int3472_clock_ops = {
 	.recalc_rate = skl_int3472_clk_recalc_rate,
 };
 
-int skl_int3472_register_clock(struct int3472_discrete_device *int3472,
-			       struct acpi_resource_gpio *agpio, u32 polarity)
+int skl_int3472_register_dsm_clock(struct int3472_discrete_device *int3472)
+{
+	struct acpi_device *adev = int3472->adev;
+	struct clk_init_data init = {
+		.ops = &skl_int3472_clock_ops,
+		.flags = CLK_GET_RATE_NOCACHE,
+	};
+	int ret;
+
+	if (int3472->clock.cl)
+		return 0; /* A GPIO controlled clk has already been registered */
+
+	if (!acpi_check_dsm(adev->handle, &img_clk_guid, 0, BIT(1)))
+		return 0; /* DSM clock control is not available */
+
+	init.name = kasprintf(GFP_KERNEL, "%s-clk", acpi_dev_name(adev));
+	if (!init.name)
+		return -ENOMEM;
+
+	int3472->clock.frequency = skl_int3472_get_clk_frequency(int3472);
+	int3472->clock.clk_hw.init = &init;
+	int3472->clock.clk = clk_register(&adev->dev, &int3472->clock.clk_hw);
+	if (IS_ERR(int3472->clock.clk)) {
+		ret = PTR_ERR(int3472->clock.clk);
+		goto out_free_init_name;
+	}
+
+	int3472->clock.cl = clkdev_create(int3472->clock.clk, NULL, int3472->sensor_name);
+	if (!int3472->clock.cl) {
+		ret = -ENOMEM;
+		goto err_unregister_clk;
+	}
+
+	kfree(init.name);
+	return 0;
+
+err_unregister_clk:
+	clk_unregister(int3472->clock.clk);
+out_free_init_name:
+	kfree(init.name);
+	return ret;
+}
+
+int skl_int3472_register_gpio_clock(struct int3472_discrete_device *int3472,
+				    struct acpi_resource_gpio *agpio, u32 polarity)
 {
 	char *path = agpio->resource_source.string_ptr;
 	struct clk_init_data init = {
diff --git a/drivers/platform/x86/intel/int3472/common.h b/drivers/platform/x86/intel/int3472/common.h
index 61688e450ce5..10a72f42a998 100644
--- a/drivers/platform/x86/intel/int3472/common.h
+++ b/drivers/platform/x86/intel/int3472/common.h
@@ -64,7 +64,9 @@ struct int3472_cldb {
 	u8 control_logic_type;
 	u8 control_logic_id;
 	u8 sensor_card_sku;
-	u8 reserved[28];
+	u8 reserved[10];
+	u8 clock_source;
+	u8 reserved2[17];
 };
 
 struct int3472_gpio_function_remap {
@@ -100,6 +102,7 @@ struct int3472_discrete_device {
 		struct clk_lookup *cl;
 		struct gpio_desc *ena_gpio;
 		u32 frequency;
+		u8 imgclk_index;
 	} clock;
 
 	struct int3472_pled {
@@ -121,8 +124,9 @@ int skl_int3472_get_sensor_adev_and_name(struct device *dev,
 					 struct acpi_device **sensor_adev_ret,
 					 const char **name_ret);
 
-int skl_int3472_register_clock(struct int3472_discrete_device *int3472,
-			       struct acpi_resource_gpio *agpio, u32 polarity);
+int skl_int3472_register_gpio_clock(struct int3472_discrete_device *int3472,
+				    struct acpi_resource_gpio *agpio, u32 polarity);
+int skl_int3472_register_dsm_clock(struct int3472_discrete_device *int3472);
 void skl_int3472_unregister_clock(struct int3472_discrete_device *int3472);
 
 int skl_int3472_register_regulator(struct int3472_discrete_device *int3472,
diff --git a/drivers/platform/x86/intel/int3472/discrete.c b/drivers/platform/x86/intel/int3472/discrete.c
index ef020e23e596..8111579a59d4 100644
--- a/drivers/platform/x86/intel/int3472/discrete.c
+++ b/drivers/platform/x86/intel/int3472/discrete.c
@@ -258,7 +258,7 @@ static int skl_int3472_handle_gpio_resources(struct acpi_resource *ares,
 
 		break;
 	case INT3472_GPIO_TYPE_CLK_ENABLE:
-		ret = skl_int3472_register_clock(int3472, agpio, polarity);
+		ret = skl_int3472_register_gpio_clock(int3472, agpio, polarity);
 		if (ret)
 			err_msg = "Failed to register clock\n";
 
@@ -311,6 +311,11 @@ static int skl_int3472_parse_crs(struct int3472_discrete_device *int3472)
 
 	acpi_dev_free_resource_list(&resource_list);
 
+	/* Register _DSM based clock (no-op if a GPIO clock was already registered) */
+	ret = skl_int3472_register_dsm_clock(int3472);
+	if (ret < 0)
+		return ret;
+
 	int3472->gpios.dev_id = int3472->sensor_name;
 	gpiod_add_lookup_table(&int3472->gpios);
 
@@ -356,6 +361,7 @@ static int skl_int3472_discrete_probe(struct platform_device *pdev)
 	int3472->adev = adev;
 	int3472->dev = &pdev->dev;
 	platform_set_drvdata(pdev, int3472);
+	int3472->clock.imgclk_index = cldb.clock_source;
 
 	ret = skl_int3472_get_sensor_adev_and_name(&pdev->dev, &int3472->sensor,
 						   &int3472->sensor_name);
-- 
2.40.1

Re: [PATCH v2] platform/x86: int3472: Evaluate device's _DSM method to control imaging clock

[Dan Scally]

Hello Hans and Bingbu (and thanks for both versions of the patch)

On 31/05/2023 14:44, Hans de Goede wrote:
> From: Bingbu Cao <bingbu.cao@intel.com>
>
> On some platforms, the imaging clock should be controlled by evaluating
> specific clock device's _DSM method instead of setting gpio, so this
> change register clock if no gpio based clock and then use the _DSM method
> to enable and disable clock.


Interesting - is that a common thing? Are there other camera-related resources that are controlled 
in a similar way? I still don't know how to drive the infrared LED on most Surface platforms, so now 
I'm wondering if they need something similar doing.


It does seem a bit strange for this to be a _DSM method against the INT3472 rather than part of _PS0 
against the camera itself - isn't that where you'd usually find such things?

>
> Signed-off-by: Bingbu Cao <bingbu.cao@intel.com>
> Signed-off-by: Hao Yao <hao.yao@intel.com>
> Link: https://lore.kernel.org/r/20230524035135.90315-2-bingbu.cao@intel.com
> ---
> Changes in v2 (Hans de Goede):
> - Minor comment / code changes (address Andy's review remarks)
> - Add a acpi_check_dsm() call
> - Return 0 instead of error if we already have a GPIO clk or if
>    acpi_check_dsm() fails
> - Rename skl_int3472_register_clock() -> skl_int3472_register_gpio_clock()
>    and name the new function: skl_int3472_register_dsm_clock()
> - Move the skl_int3472_register_dsm_clock() call to after
>    acpi_dev_free_resource_list() and add error checking for it


I think all these changes are good ones.

> ---
>   .../x86/intel/int3472/clk_and_regulator.c     | 89 ++++++++++++++++++-
>   drivers/platform/x86/intel/int3472/common.h   | 10 ++-
>   drivers/platform/x86/intel/int3472/discrete.c |  8 +-
>   3 files changed, 99 insertions(+), 8 deletions(-)
>
> diff --git a/drivers/platform/x86/intel/int3472/clk_and_regulator.c b/drivers/platform/x86/intel/int3472/clk_and_regulator.c
> index 1086c3d83494..9bcf8c64b8e7 100644
> --- a/drivers/platform/x86/intel/int3472/clk_and_regulator.c
> +++ b/drivers/platform/x86/intel/int3472/clk_and_regulator.c
> @@ -11,6 +11,37 @@
>   
>   #include "common.h"
>   
> +/*
> + * 82c0d13a-78c5-4244-9bb1-eb8b539a8d11
> + * This _DSM GUID allows controlling the sensor clk when it is not controlled
> + * through a GPIO.
> + */
> +static const guid_t img_clk_guid =
> +	GUID_INIT(0x82c0d13a, 0x78c5, 0x4244,
> +		  0x9b, 0xb1, 0xeb, 0x8b, 0x53, 0x9a, 0x8d, 0x11);
> +
> +static void skl_int3472_enable_clk_acpi_method(struct int3472_gpio_clock *clk,
> +					       bool enable)
> +{
> +	struct int3472_discrete_device *int3472 = to_int3472_device(clk);
> +	union acpi_object args[3];
> +	union acpi_object argv4;
> +
> +	args[0].integer.type = ACPI_TYPE_INTEGER;
> +	args[0].integer.value = clk->imgclk_index;
> +	args[1].integer.type = ACPI_TYPE_INTEGER;
> +	args[1].integer.value = enable ? 1 : 0;
> +	args[2].integer.type = ACPI_TYPE_INTEGER;
> +	args[2].integer.value = 1;
> +
> +	argv4.type = ACPI_TYPE_PACKAGE;
> +	argv4.package.count = 3;
> +	argv4.package.elements = args;
> +
> +	acpi_evaluate_dsm(acpi_device_handle(int3472->adev), &img_clk_guid,
> +			  0, 1, &argv4);
> +}


I'm not really sure what error modes something like this might have, but acpi_evaluate_dsm() at 
least can fail - is there no value in error checking here so that it could be returned by 
skl_int3472_clk_prepare() below?

> +
>   /*
>    * The regulators have to have .ops to be valid, but the only ops we actually
>    * support are .enable and .disable which are handled via .ena_gpiod. Pass an
> @@ -22,7 +53,11 @@ static int skl_int3472_clk_prepare(struct clk_hw *hw)
>   {
>   	struct int3472_gpio_clock *clk = to_int3472_clk(hw);
>   
> -	gpiod_set_value_cansleep(clk->ena_gpio, 1);
> +	if (clk->ena_gpio)
> +		gpiod_set_value_cansleep(clk->ena_gpio, 1);
> +	else
> +		skl_int3472_enable_clk_acpi_method(clk, 1);
> +
>   	return 0;
>   }
>   
> @@ -30,7 +65,10 @@ static void skl_int3472_clk_unprepare(struct clk_hw *hw)
>   {
>   	struct int3472_gpio_clock *clk = to_int3472_clk(hw);
>   
> -	gpiod_set_value_cansleep(clk->ena_gpio, 0);
> +	if (clk->ena_gpio)
> +		gpiod_set_value_cansleep(clk->ena_gpio, 0);
> +	else
> +		skl_int3472_enable_clk_acpi_method(clk, 0);
>   }
>   
>   static int skl_int3472_clk_enable(struct clk_hw *hw)
> @@ -86,8 +124,51 @@ static const struct clk_ops skl_int3472_clock_ops = {
>   	.recalc_rate = skl_int3472_clk_recalc_rate,
>   };
>   
> -int skl_int3472_register_clock(struct int3472_discrete_device *int3472,
> -			       struct acpi_resource_gpio *agpio, u32 polarity)
> +int skl_int3472_register_dsm_clock(struct int3472_discrete_device *int3472)
> +{
> +	struct acpi_device *adev = int3472->adev;
> +	struct clk_init_data init = {
> +		.ops = &skl_int3472_clock_ops,
> +		.flags = CLK_GET_RATE_NOCACHE,
> +	};
> +	int ret;
> +
> +	if (int3472->clock.cl)
> +		return 0; /* A GPIO controlled clk has already been registered */
> +
> +	if (!acpi_check_dsm(adev->handle, &img_clk_guid, 0, BIT(1)))
> +		return 0; /* DSM clock control is not available */
> +
> +	init.name = kasprintf(GFP_KERNEL, "%s-clk", acpi_dev_name(adev));
> +	if (!init.name)
> +		return -ENOMEM;
> +
> +	int3472->clock.frequency = skl_int3472_get_clk_frequency(int3472);
> +	int3472->clock.clk_hw.init = &init;
> +	int3472->clock.clk = clk_register(&adev->dev, &int3472->clock.clk_hw);
> +	if (IS_ERR(int3472->clock.clk)) {
> +		ret = PTR_ERR(int3472->clock.clk);
> +		goto out_free_init_name;
> +	}
> +
> +	int3472->clock.cl = clkdev_create(int3472->clock.clk, NULL, int3472->sensor_name);
> +	if (!int3472->clock.cl) {
> +		ret = -ENOMEM;
> +		goto err_unregister_clk;
> +	}
> +
> +	kfree(init.name);
> +	return 0;
> +
> +err_unregister_clk:
> +	clk_unregister(int3472->clock.clk);
> +out_free_init_name:
> +	kfree(init.name);
> +	return ret;
> +}
> +
> +int skl_int3472_register_gpio_clock(struct int3472_discrete_device *int3472,
> +				    struct acpi_resource_gpio *agpio, u32 polarity)
>   {
>   	char *path = agpio->resource_source.string_ptr;
>   	struct clk_init_data init = {
> diff --git a/drivers/platform/x86/intel/int3472/common.h b/drivers/platform/x86/intel/int3472/common.h
> index 61688e450ce5..10a72f42a998 100644
> --- a/drivers/platform/x86/intel/int3472/common.h
> +++ b/drivers/platform/x86/intel/int3472/common.h
> @@ -64,7 +64,9 @@ struct int3472_cldb {
>   	u8 control_logic_type;
>   	u8 control_logic_id;
>   	u8 sensor_card_sku;
> -	u8 reserved[28];
> +	u8 reserved[10];
> +	u8 clock_source;
> +	u8 reserved2[17];


Not really a comment on the functionality of the patch, but can we not just get the rest of those 
fields filled out?

>   };
>   
>   struct int3472_gpio_function_remap {
> @@ -100,6 +102,7 @@ struct int3472_discrete_device {
>   		struct clk_lookup *cl;
>   		struct gpio_desc *ena_gpio;
>   		u32 frequency;
> +		u8 imgclk_index;
>   	} clock;


This struct is called "int3472_gpio_clock" but perhaps now ought to just be "int3472_clock"

>   
>   	struct int3472_pled {
> @@ -121,8 +124,9 @@ int skl_int3472_get_sensor_adev_and_name(struct device *dev,
>   					 struct acpi_device **sensor_adev_ret,
>   					 const char **name_ret);
>   
> -int skl_int3472_register_clock(struct int3472_discrete_device *int3472,
> -			       struct acpi_resource_gpio *agpio, u32 polarity);
> +int skl_int3472_register_gpio_clock(struct int3472_discrete_device *int3472,
> +				    struct acpi_resource_gpio *agpio, u32 polarity);
> +int skl_int3472_register_dsm_clock(struct int3472_discrete_device *int3472);
>   void skl_int3472_unregister_clock(struct int3472_discrete_device *int3472);
>   
>   int skl_int3472_register_regulator(struct int3472_discrete_device *int3472,
> diff --git a/drivers/platform/x86/intel/int3472/discrete.c b/drivers/platform/x86/intel/int3472/discrete.c
> index ef020e23e596..8111579a59d4 100644
> --- a/drivers/platform/x86/intel/int3472/discrete.c
> +++ b/drivers/platform/x86/intel/int3472/discrete.c
> @@ -258,7 +258,7 @@ static int skl_int3472_handle_gpio_resources(struct acpi_resource *ares,
>   
>   		break;
>   	case INT3472_GPIO_TYPE_CLK_ENABLE:
> -		ret = skl_int3472_register_clock(int3472, agpio, polarity);
> +		ret = skl_int3472_register_gpio_clock(int3472, agpio, polarity);
>   		if (ret)
>   			err_msg = "Failed to register clock\n";
>   
> @@ -311,6 +311,11 @@ static int skl_int3472_parse_crs(struct int3472_discrete_device *int3472)
>   
>   	acpi_dev_free_resource_list(&resource_list);
>   
> +	/* Register _DSM based clock (no-op if a GPIO clock was already registered) */
> +	ret = skl_int3472_register_dsm_clock(int3472);
> +	if (ret < 0)
> +		return ret;
> +
>   	int3472->gpios.dev_id = int3472->sensor_name;
>   	gpiod_add_lookup_table(&int3472->gpios);
>   
> @@ -356,6 +361,7 @@ static int skl_int3472_discrete_probe(struct platform_device *pdev)
>   	int3472->adev = adev;
>   	int3472->dev = &pdev->dev;
>   	platform_set_drvdata(pdev, int3472);
> +	int3472->clock.imgclk_index = cldb.clock_source;
>   
>   	ret = skl_int3472_get_sensor_adev_and_name(&pdev->dev, &int3472->sensor,
>   						   &int3472->sensor_name);

Re: [PATCH v2] platform/x86: int3472: Evaluate device's _DSM method to control imaging clock

[Hans de Goede]

Hi Dan,

On 5/31/23 16:45, Dan Scally wrote:
> Hello Hans and Bingbu (and thanks for both versions of the patch)
> 
> On 31/05/2023 14:44, Hans de Goede wrote:
>> From: Bingbu Cao <bingbu.cao@intel.com>
>>
>> On some platforms, the imaging clock should be controlled by evaluating
>> specific clock device's _DSM method instead of setting gpio, so this
>> change register clock if no gpio based clock and then use the _DSM method
>> to enable and disable clock.
> 
> 
> Interesting - is that a common thing? Are there other camera-related resources that are controlled in a similar way? I still don't know how to drive the infrared LED on most Surface platforms, so now I'm wondering if they need something similar doing.
> 
> 
> It does seem a bit strange for this to be a _DSM method against the INT3472 rather than part of _PS0 against the camera itself - isn't that where you'd usually find such things?
> 
>>
>> Signed-off-by: Bingbu Cao <bingbu.cao@intel.com>
>> Signed-off-by: Hao Yao <hao.yao@intel.com>
>> Link: https://lore.kernel.org/r/20230524035135.90315-2-bingbu.cao@intel.com
>> ---
>> Changes in v2 (Hans de Goede):
>> - Minor comment / code changes (address Andy's review remarks)
>> - Add a acpi_check_dsm() call
>> - Return 0 instead of error if we already have a GPIO clk or if
>>    acpi_check_dsm() fails
>> - Rename skl_int3472_register_clock() -> skl_int3472_register_gpio_clock()
>>    and name the new function: skl_int3472_register_dsm_clock()
>> - Move the skl_int3472_register_dsm_clock() call to after
>>    acpi_dev_free_resource_list() and add error checking for it
> 
> 
> I think all these changes are good ones.
> 
>> ---
>>   .../x86/intel/int3472/clk_and_regulator.c     | 89 ++++++++++++++++++-
>>   drivers/platform/x86/intel/int3472/common.h   | 10 ++-
>>   drivers/platform/x86/intel/int3472/discrete.c |  8 +-
>>   3 files changed, 99 insertions(+), 8 deletions(-)
>>
>> diff --git a/drivers/platform/x86/intel/int3472/clk_and_regulator.c b/drivers/platform/x86/intel/int3472/clk_and_regulator.c
>> index 1086c3d83494..9bcf8c64b8e7 100644
>> --- a/drivers/platform/x86/intel/int3472/clk_and_regulator.c
>> +++ b/drivers/platform/x86/intel/int3472/clk_and_regulator.c
>> @@ -11,6 +11,37 @@
>>     #include "common.h"
>>   +/*
>> + * 82c0d13a-78c5-4244-9bb1-eb8b539a8d11
>> + * This _DSM GUID allows controlling the sensor clk when it is not controlled
>> + * through a GPIO.
>> + */
>> +static const guid_t img_clk_guid =
>> +    GUID_INIT(0x82c0d13a, 0x78c5, 0x4244,
>> +          0x9b, 0xb1, 0xeb, 0x8b, 0x53, 0x9a, 0x8d, 0x11);
>> +
>> +static void skl_int3472_enable_clk_acpi_method(struct int3472_gpio_clock *clk,
>> +                           bool enable)
>> +{
>> +    struct int3472_discrete_device *int3472 = to_int3472_device(clk);
>> +    union acpi_object args[3];
>> +    union acpi_object argv4;
>> +
>> +    args[0].integer.type = ACPI_TYPE_INTEGER;
>> +    args[0].integer.value = clk->imgclk_index;
>> +    args[1].integer.type = ACPI_TYPE_INTEGER;
>> +    args[1].integer.value = enable ? 1 : 0;
>> +    args[2].integer.type = ACPI_TYPE_INTEGER;
>> +    args[2].integer.value = 1;
>> +
>> +    argv4.type = ACPI_TYPE_PACKAGE;
>> +    argv4.package.count = 3;
>> +    argv4.package.elements = args;
>> +
>> +    acpi_evaluate_dsm(acpi_device_handle(int3472->adev), &img_clk_guid,
>> +              0, 1, &argv4);
>> +}
> 
> 
> I'm not really sure what error modes something like this might have, but acpi_evaluate_dsm() at least can fail - is there no value in error checking here so that it could be returned by skl_int3472_clk_prepare() below?

The problem is that acpi_evaluate_dsm() returns an acpi_object * not a status.

And it returns NULL on errors as well as if the _DSM method returns nothing (ACPI
equivalent of returning void).

And these clk-set calls are expected to return void.

The good news is that acpi_evaluate_dsm() does log an error one errors.

<snip>

>> @@ -100,6 +102,7 @@ struct int3472_discrete_device {
>>           struct clk_lookup *cl;
>>           struct gpio_desc *ena_gpio;
>>           u32 frequency;
>> +        u8 imgclk_index;
>>       } clock;
> 
> 
> This struct is called "int3472_gpio_clock" but perhaps now ought to just be "int3472_clock"

Ack I've fixed this up while merging the patch.

Regards,

Hans

Re: [PATCH v2] platform/x86: int3472: Evaluate device's _DSM method to control imaging clock

[Hans de Goede]

HI,

On 5/31/23 19:56, Andy Shevchenko wrote:
> On Wed, May 31, 2023 at 4:44 PM Hans de Goede <hdegoede@redhat.com> wrote:
>>
>> From: Bingbu Cao <bingbu.cao@intel.com>
>>
>> On some platforms, the imaging clock should be controlled by evaluating
>> specific clock device's _DSM method instead of setting gpio, so this
>> change register clock if no gpio based clock and then use the _DSM method
>> to enable and disable clock.
> 
> ...
> 
>> +       if (clk->ena_gpio)
>> +               gpiod_set_value_cansleep(clk->ena_gpio, 1);
>> +       else
>> +               skl_int3472_enable_clk_acpi_method(clk, 1);
> 
> Looking at this, can we avoid duplicative validation of the GPIO?
> Perhaps skl_int3472_enable_clk_acpi_method() can have embedded another
> check so they won't be called together?
> 
> ...
> 
>> +       if (clk->ena_gpio)
>> +               gpiod_set_value_cansleep(clk->ena_gpio, 0);
>> +       else
>> +               skl_int3472_enable_clk_acpi_method(clk, 0);
> 
> Ditto.

Ack, I've squashed a fix for this into this patch while merging it into
my review-hans branch.

Regards,

Hans

Re: [PATCH v2] platform/x86: int3472: Evaluate device's _DSM method to control imaging clock

[Andy Shevchenko]

On Tue, Jun 6, 2023 at 12:23 PM Hans de Goede <hdegoede@redhat.com> wrote:
> On 5/31/23 19:56, Andy Shevchenko wrote:
> > On Wed, May 31, 2023 at 4:44 PM Hans de Goede <hdegoede@redhat.com> wrote:

...

> >> +       if (clk->ena_gpio)
> >> +               gpiod_set_value_cansleep(clk->ena_gpio, 1);
> >> +       else
> >> +               skl_int3472_enable_clk_acpi_method(clk, 1);
> >
> > Looking at this, can we avoid duplicative validation of the GPIO?
> > Perhaps skl_int3472_enable_clk_acpi_method() can have embedded another
> > check so they won't be called together?
> >
> > ...
> >
> >> +       if (clk->ena_gpio)
> >> +               gpiod_set_value_cansleep(clk->ena_gpio, 0);
> >> +       else
> >> +               skl_int3472_enable_clk_acpi_method(clk, 0);
> >
> > Ditto.
>
> Ack, I've squashed a fix for this into this patch while merging it into
> my review-hans branch.

Now you have two different checks for the same, I would suggest that
you switch to check clk->cl in the skl_int3472_enable_clk()
as it's done in the skl_int3472_register_dsm_clock() instead of GPIO.
Other way around is also possible but it seems to me that checking for
clock existence has better guarantees than just checking for GPIO
availability.

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH v2] platform/x86: int3472: Evaluate device's _DSM method to control imaging clock

[Andy Shevchenko]

On Tue, Jun 6, 2023 at 2:26 PM Andy Shevchenko
<andy.shevchenko@gmail.com> wrote:
> On Tue, Jun 6, 2023 at 12:23 PM Hans de Goede <hdegoede@redhat.com> wrote:
> > On 5/31/23 19:56, Andy Shevchenko wrote:

...

> > Ack, I've squashed a fix for this into this patch while merging it into
> > my review-hans branch.
>
> Now you have two different checks for the same, I would suggest that
> you switch to check clk->cl in the skl_int3472_enable_clk()
> as it's done in the skl_int3472_register_dsm_clock() instead of GPIO.
> Other way around is also possible but it seems to me that checking for
> clock existence has better guarantees than just checking for GPIO
> availability.

That said it might make sense to also introduce

struct ... *clk = &int3472->clock;

in skl_int3472_register_dsm_clock() so the above mentioned checks will
be the same and actually code in the latter will look neater (?).


-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH v2] platform/x86: int3472: Evaluate device's _DSM method to control imaging clock

[Hans de Goede]

Hi,

On 6/6/23 13:26, Andy Shevchenko wrote:
> On Tue, Jun 6, 2023 at 12:23 PM Hans de Goede <hdegoede@redhat.com> wrote:
>> On 5/31/23 19:56, Andy Shevchenko wrote:
>>> On Wed, May 31, 2023 at 4:44 PM Hans de Goede <hdegoede@redhat.com> wrote:
> 
> ...
> 
>>>> +       if (clk->ena_gpio)
>>>> +               gpiod_set_value_cansleep(clk->ena_gpio, 1);
>>>> +       else
>>>> +               skl_int3472_enable_clk_acpi_method(clk, 1);
>>>
>>> Looking at this, can we avoid duplicative validation of the GPIO?
>>> Perhaps skl_int3472_enable_clk_acpi_method() can have embedded another
>>> check so they won't be called together?
>>>
>>> ...
>>>
>>>> +       if (clk->ena_gpio)
>>>> +               gpiod_set_value_cansleep(clk->ena_gpio, 0);
>>>> +       else
>>>> +               skl_int3472_enable_clk_acpi_method(clk, 0);
>>>
>>> Ditto.
>>
>> Ack, I've squashed a fix for this into this patch while merging it into
>> my review-hans branch.
> 
> Now you have two different checks for the same, I would suggest that
> you switch to check clk->cl in the skl_int3472_enable_clk()
> as it's done in the skl_int3472_register_dsm_clock() instead of GPIO.
> Other way around is also possible but it seems to me that checking for
> clock existence has better guarantees than just checking for GPIO
> availability.

Hmm, you mean the:

        if (int3472->clock.cl)
                return 0; /* A GPIO controlled clk has already been registered */

Check in skl_int3472_register_dsm_clock() ? That matches with / aligns with
this check:

        if (int3472->clock.cl)
                return -EBUSY;

in skl_int3472_register_gpio_clock().

To me it seems sensible to align the checks for "a clk has already been
registered" up this way.

Regards,

Hans

Re: [PATCH v2] platform/x86: int3472: Evaluate device's _DSM method to control imaging clock

[Andy Shevchenko]

On Tue, Jun 06, 2023 at 03:13:47PM +0200, Hans de Goede wrote:
> On 6/6/23 13:26, Andy Shevchenko wrote:
> > On Tue, Jun 6, 2023 at 12:23 PM Hans de Goede <hdegoede@redhat.com> wrote:
> >> On 5/31/23 19:56, Andy Shevchenko wrote:
> >>> On Wed, May 31, 2023 at 4:44 PM Hans de Goede <hdegoede@redhat.com> wrote:

...

> >>>> +       if (clk->ena_gpio)
> >>>> +               gpiod_set_value_cansleep(clk->ena_gpio, 1);
> >>>> +       else
> >>>> +               skl_int3472_enable_clk_acpi_method(clk, 1);
> >>>
> >>> Looking at this, can we avoid duplicative validation of the GPIO?
> >>> Perhaps skl_int3472_enable_clk_acpi_method() can have embedded another
> >>> check so they won't be called together?
> >>>
> >>> ...
> >>>
> >>>> +       if (clk->ena_gpio)
> >>>> +               gpiod_set_value_cansleep(clk->ena_gpio, 0);
> >>>> +       else
> >>>> +               skl_int3472_enable_clk_acpi_method(clk, 0);
> >>>
> >>> Ditto.
> >>
> >> Ack, I've squashed a fix for this into this patch while merging it into
> >> my review-hans branch.
> > 
> > Now you have two different checks for the same, I would suggest that
> > you switch to check clk->cl in the skl_int3472_enable_clk()
> > as it's done in the skl_int3472_register_dsm_clock() instead of GPIO.
> > Other way around is also possible but it seems to me that checking for
> > clock existence has better guarantees than just checking for GPIO
> > availability.
> 
> Hmm, you mean the:
> 
>         if (int3472->clock.cl)
>                 return 0; /* A GPIO controlled clk has already been registered */
> 
> Check in skl_int3472_register_dsm_clock() ? That matches with / aligns with
> this check:
> 
>         if (int3472->clock.cl)
>                 return -EBUSY;
> 
> in skl_int3472_register_gpio_clock().
> 
> To me it seems sensible to align the checks for "a clk has already been
> registered" up this way.

I'm talking about enable method when we actually change the clock state.

-- 
With Best Regards,
Andy Shevchenko