[PATCH v4 0/2] hwmon: scmi: Scale values to target desired HWMON units

[PATCH v4 0/2] hwmon: scmi: Scale values to target desired HWMON units

[Florian Fainelli]

Hi Sudeep, Guenter,

This patch series adds support for scaling SCMI sensor values read from
firmware. Sudeep, let me know if you think we should be treating scale
== 0 as a special value to preserve some firmware compatibility (not
that this would be desired).

Changes in v4:
- deal with overflow in the caller of __pow10() as suggested by Guenter
  which makes us rework a bit how the value argument/return value are
  passed
- don't harcode the latest power of 10 factor to be 18, just rely on
  overflowing the u64 value instead

Changes in v3:

- add a local __pow10 function to scmi-hwmon.c while a plan to provide
  a generic function is figured out.
- add check on power > 18 which would overflow a 64-bit unsigned integer
- use div64_u64() to properly divide a 64-bit quantity with an unsigned
  64-bit quantity

Changes in v2:

- added a helper function in kernel.h: __pow10()
- made the scale in scmi_sensor_info an s8 type, added defines for
  checking the sign bit and sign extending with a mask
- simplify computations in hwmon driver

Florian Fainelli (2):
  firmware: arm_scmi: Fetch and store sensor scale
  hwmon: scmi: Scale values to target desired HWMON units

 drivers/firmware/arm_scmi/sensors.c |  6 ++++
 drivers/hwmon/scmi-hwmon.c          | 46 +++++++++++++++++++++++++++++
 include/linux/scmi_protocol.h       |  1 +
 3 files changed, 53 insertions(+)

-- 
2.17.1

[PATCH v4 1/2] firmware: arm_scmi: Fetch and store sensor scale

[Florian Fainelli]

In preparation for dealing with scales within the SCMI HWMON driver,
fetch and store the sensor unit scale into the scmi_sensor_info
structure. In order to simplify computations for upper layer, take care
of sign extending the scale to a full 8-bit signed value.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
---
 drivers/firmware/arm_scmi/sensors.c | 6 ++++++
 include/linux/scmi_protocol.h       | 1 +
 2 files changed, 7 insertions(+)

diff --git a/drivers/firmware/arm_scmi/sensors.c b/drivers/firmware/arm_scmi/sensors.c
index b53d5cc9c9f6..21353470a740 100644
--- a/drivers/firmware/arm_scmi/sensors.c
+++ b/drivers/firmware/arm_scmi/sensors.c
@@ -34,6 +34,8 @@ struct scmi_msg_resp_sensor_description {
 		__le32 attributes_high;
 #define SENSOR_TYPE(x)		((x) & 0xff)
 #define SENSOR_SCALE(x)		(((x) >> 11) & 0x3f)
+#define SENSOR_SCALE_SIGN	BIT(5)
+#define SENSOR_SCALE_EXTEND	GENMASK(7, 6)
 #define SENSOR_UPDATE_SCALE(x)	(((x) >> 22) & 0x1f)
 #define SENSOR_UPDATE_BASE(x)	(((x) >> 27) & 0x1f)
 		    u8 name[SCMI_MAX_STR_SIZE];
@@ -140,6 +142,10 @@ static int scmi_sensor_description_get(const struct scmi_handle *handle,
 			s = &si->sensors[desc_index + cnt];
 			s->id = le32_to_cpu(buf->desc[cnt].id);
 			s->type = SENSOR_TYPE(attrh);
+			s->scale = SENSOR_SCALE(attrh);
+			/* Sign extend to a full s8 */
+			if (s->scale & SENSOR_SCALE_SIGN)
+				s->scale |= SENSOR_SCALE_EXTEND;
 			strlcpy(s->name, buf->desc[cnt].name, SCMI_MAX_STR_SIZE);
 		}
 
diff --git a/include/linux/scmi_protocol.h b/include/linux/scmi_protocol.h
index 3105055c00a7..9ff2e9357e9a 100644
--- a/include/linux/scmi_protocol.h
+++ b/include/linux/scmi_protocol.h
@@ -144,6 +144,7 @@ struct scmi_power_ops {
 struct scmi_sensor_info {
 	u32 id;
 	u8 type;
+	s8 scale;
 	char name[SCMI_MAX_STR_SIZE];
 };
 
-- 
2.17.1

Re: [PATCH v4 1/2] firmware: arm_scmi: Fetch and store sensor scale

[Guenter Roeck]

On Wed, May 08, 2019 at 10:00:34AM -0700, Florian Fainelli wrote:
> In preparation for dealing with scales within the SCMI HWMON driver,
> fetch and store the sensor unit scale into the scmi_sensor_info
> structure. In order to simplify computations for upper layer, take care
> of sign extending the scale to a full 8-bit signed value.
> 
> Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>

Reviewed-by: Guenter Roeck <linux@roeck-us.net>

> ---
>  drivers/firmware/arm_scmi/sensors.c | 6 ++++++
>  include/linux/scmi_protocol.h       | 1 +
>  2 files changed, 7 insertions(+)
> 
> diff --git a/drivers/firmware/arm_scmi/sensors.c b/drivers/firmware/arm_scmi/sensors.c
> index b53d5cc9c9f6..21353470a740 100644
> --- a/drivers/firmware/arm_scmi/sensors.c
> +++ b/drivers/firmware/arm_scmi/sensors.c
> @@ -34,6 +34,8 @@ struct scmi_msg_resp_sensor_description {
>  		__le32 attributes_high;
>  #define SENSOR_TYPE(x)		((x) & 0xff)
>  #define SENSOR_SCALE(x)		(((x) >> 11) & 0x3f)
> +#define SENSOR_SCALE_SIGN	BIT(5)
> +#define SENSOR_SCALE_EXTEND	GENMASK(7, 6)
>  #define SENSOR_UPDATE_SCALE(x)	(((x) >> 22) & 0x1f)
>  #define SENSOR_UPDATE_BASE(x)	(((x) >> 27) & 0x1f)
>  		    u8 name[SCMI_MAX_STR_SIZE];
> @@ -140,6 +142,10 @@ static int scmi_sensor_description_get(const struct scmi_handle *handle,
>  			s = &si->sensors[desc_index + cnt];
>  			s->id = le32_to_cpu(buf->desc[cnt].id);
>  			s->type = SENSOR_TYPE(attrh);
> +			s->scale = SENSOR_SCALE(attrh);
> +			/* Sign extend to a full s8 */
> +			if (s->scale & SENSOR_SCALE_SIGN)
> +				s->scale |= SENSOR_SCALE_EXTEND;
>  			strlcpy(s->name, buf->desc[cnt].name, SCMI_MAX_STR_SIZE);
>  		}
>  
> diff --git a/include/linux/scmi_protocol.h b/include/linux/scmi_protocol.h
> index 3105055c00a7..9ff2e9357e9a 100644
> --- a/include/linux/scmi_protocol.h
> +++ b/include/linux/scmi_protocol.h
> @@ -144,6 +144,7 @@ struct scmi_power_ops {
>  struct scmi_sensor_info {
>  	u32 id;
>  	u8 type;
> +	s8 scale;
>  	char name[SCMI_MAX_STR_SIZE];
>  };
>  
> -- 
> 2.17.1
> 

[PATCH v4 2/2] hwmon: scmi: Scale values to target desired HWMON units

[Florian Fainelli]

If the SCMI firmware implementation is reporting values in a scale that
is different from the HWMON units, we need to scale up or down the value
according to how far appart they are.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
---
 drivers/hwmon/scmi-hwmon.c | 46 ++++++++++++++++++++++++++++++++++++++
 1 file changed, 46 insertions(+)

diff --git a/drivers/hwmon/scmi-hwmon.c b/drivers/hwmon/scmi-hwmon.c
index a80183a488c5..4399372e2131 100644
--- a/drivers/hwmon/scmi-hwmon.c
+++ b/drivers/hwmon/scmi-hwmon.c
@@ -7,6 +7,7 @@
  */
 
 #include <linux/hwmon.h>
+#include <linux/limits.h>
 #include <linux/module.h>
 #include <linux/scmi_protocol.h>
 #include <linux/slab.h>
@@ -18,6 +19,47 @@ struct scmi_sensors {
 	const struct scmi_sensor_info **info[hwmon_max];
 };
 
+static inline u64 __pow10(u8 x)
+{
+	u64 r = 1;
+
+	while (x--)
+		r *= 10;
+
+	return r;
+}
+
+static int scmi_hwmon_scale(const struct scmi_sensor_info *sensor, u64 *value)
+{
+	s8 scale = sensor->scale;
+	u64 f;
+
+	switch (sensor->type) {
+	case TEMPERATURE_C:
+	case VOLTAGE:
+	case CURRENT:
+		scale += 3;
+		break;
+	case POWER:
+	case ENERGY:
+		scale += 6;
+		break;
+	default:
+		break;
+	}
+
+	f = __pow10(abs(scale));
+	if (f == U64_MAX)
+		return -E2BIG;
+
+	if (scale > 0)
+		*value *= f;
+	else
+		*value = div64_u64(*value, f);
+
+        return 0;
+}
+
 static int scmi_hwmon_read(struct device *dev, enum hwmon_sensor_types type,
 			   u32 attr, int channel, long *val)
 {
@@ -29,6 +71,10 @@ static int scmi_hwmon_read(struct device *dev, enum hwmon_sensor_types type,
 
 	sensor = *(scmi_sensors->info[type] + channel);
 	ret = h->sensor_ops->reading_get(h, sensor->id, false, &value);
+	if (ret)
+		return ret;
+
+	ret = scmi_hwmon_scale(sensor, value);
 	if (!ret)
 		*val = value;
 
-- 
2.17.1

Re: [PATCH v4 2/2] hwmon: scmi: Scale values to target desired HWMON units

[Guenter Roeck]

Hi Florian,

On Wed, May 08, 2019 at 10:00:35AM -0700, Florian Fainelli wrote:
> If the SCMI firmware implementation is reporting values in a scale that
> is different from the HWMON units, we need to scale up or down the value
> according to how far appart they are.
> 
> Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
> ---
>  drivers/hwmon/scmi-hwmon.c | 46 ++++++++++++++++++++++++++++++++++++++
>  1 file changed, 46 insertions(+)
> 
> diff --git a/drivers/hwmon/scmi-hwmon.c b/drivers/hwmon/scmi-hwmon.c
> index a80183a488c5..4399372e2131 100644
> --- a/drivers/hwmon/scmi-hwmon.c
> +++ b/drivers/hwmon/scmi-hwmon.c
> @@ -7,6 +7,7 @@
>   */
>  
>  #include <linux/hwmon.h>
> +#include <linux/limits.h>
>  #include <linux/module.h>
>  #include <linux/scmi_protocol.h>
>  #include <linux/slab.h>
> @@ -18,6 +19,47 @@ struct scmi_sensors {
>  	const struct scmi_sensor_info **info[hwmon_max];
>  };
>  
> +static inline u64 __pow10(u8 x)
> +{
> +	u64 r = 1;
> +
> +	while (x--)
> +		r *= 10;
> +
> +	return r;
> +}
> +
> +static int scmi_hwmon_scale(const struct scmi_sensor_info *sensor, u64 *value)
> +{
> +	s8 scale = sensor->scale;
> +	u64 f;
> +
> +	switch (sensor->type) {
> +	case TEMPERATURE_C:
> +	case VOLTAGE:
> +	case CURRENT:
> +		scale += 3;
> +		break;
> +	case POWER:
> +	case ENERGY:
> +		scale += 6;
> +		break;
> +	default:
> +		break;
> +	}
> +
> +	f = __pow10(abs(scale));
> +	if (f == U64_MAX)
> +		return -E2BIG;

Unfortunately that is not how integer overflows work.

A test program with increasing values of scale reports:

0: 1
...
18: 1000000000000000000
19: 10000000000000000000
20: 7766279631452241920
21: 3875820019684212736
22: 1864712049423024128
23: 200376420520689664
24: 2003764205206896640
...
61: 11529215046068469760
62: 4611686018427387904
63: 9223372036854775808
64: 0
...

You'll have to check for abs(scale) > 19 if you want to report overflows.

Guenter

> +
> +	if (scale > 0)
> +		*value *= f;
> +	else
> +		*value = div64_u64(*value, f);
> +
> +        return 0;
> +}
> +
>  static int scmi_hwmon_read(struct device *dev, enum hwmon_sensor_types type,
>  			   u32 attr, int channel, long *val)
>  {
> @@ -29,6 +71,10 @@ static int scmi_hwmon_read(struct device *dev, enum hwmon_sensor_types type,
>  
>  	sensor = *(scmi_sensors->info[type] + channel);
>  	ret = h->sensor_ops->reading_get(h, sensor->id, false, &value);
> +	if (ret)
> +		return ret;
> +
> +	ret = scmi_hwmon_scale(sensor, value);
>  	if (!ret)
>  		*val = value;
>  
> -- 
> 2.17.1
> 

Re: [PATCH v4 2/2] hwmon: scmi: Scale values to target desired HWMON units

[Florian Fainelli]

On 5/8/19 11:32 AM, Guenter Roeck wrote:
> Hi Florian,
> 
> On Wed, May 08, 2019 at 10:00:35AM -0700, Florian Fainelli wrote:
>> If the SCMI firmware implementation is reporting values in a scale that
>> is different from the HWMON units, we need to scale up or down the value
>> according to how far appart they are.
>>
>> Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
>> ---
>>  drivers/hwmon/scmi-hwmon.c | 46 ++++++++++++++++++++++++++++++++++++++
>>  1 file changed, 46 insertions(+)
>>
>> diff --git a/drivers/hwmon/scmi-hwmon.c b/drivers/hwmon/scmi-hwmon.c
>> index a80183a488c5..4399372e2131 100644
>> --- a/drivers/hwmon/scmi-hwmon.c
>> +++ b/drivers/hwmon/scmi-hwmon.c
>> @@ -7,6 +7,7 @@
>>   */
>>  
>>  #include <linux/hwmon.h>
>> +#include <linux/limits.h>
>>  #include <linux/module.h>
>>  #include <linux/scmi_protocol.h>
>>  #include <linux/slab.h>
>> @@ -18,6 +19,47 @@ struct scmi_sensors {
>>  	const struct scmi_sensor_info **info[hwmon_max];
>>  };
>>  
>> +static inline u64 __pow10(u8 x)
>> +{
>> +	u64 r = 1;
>> +
>> +	while (x--)
>> +		r *= 10;
>> +
>> +	return r;
>> +}
>> +
>> +static int scmi_hwmon_scale(const struct scmi_sensor_info *sensor, u64 *value)
>> +{
>> +	s8 scale = sensor->scale;
>> +	u64 f;
>> +
>> +	switch (sensor->type) {
>> +	case TEMPERATURE_C:
>> +	case VOLTAGE:
>> +	case CURRENT:
>> +		scale += 3;
>> +		break;
>> +	case POWER:
>> +	case ENERGY:
>> +		scale += 6;
>> +		break;
>> +	default:
>> +		break;
>> +	}
>> +
>> +	f = __pow10(abs(scale));
>> +	if (f == U64_MAX)
>> +		return -E2BIG;
> 
> Unfortunately that is not how integer overflows work.
> 
> A test program with increasing values of scale reports:
> 
> 0: 1
> ...
> 18: 1000000000000000000
> 19: 10000000000000000000
> 20: 7766279631452241920
> 21: 3875820019684212736
> 22: 1864712049423024128
> 23: 200376420520689664
> 24: 2003764205206896640
> ...
> 61: 11529215046068469760
> 62: 4611686018427387904
> 63: 9223372036854775808
> 64: 0
> ...
> 
> You'll have to check for abs(scale) > 19 if you want to report overflows.

Yes silly me, my test program was flawed, thanks for pointing out that.
You are okay with returning E2BIG when we overflow?
-- 
Florian

Re: [PATCH v4 2/2] hwmon: scmi: Scale values to target desired HWMON units

[Guenter Roeck]

On Wed, May 08, 2019 at 11:34:50AM -0700, Florian Fainelli wrote:
> On 5/8/19 11:32 AM, Guenter Roeck wrote:
> > Hi Florian,
> > 
> > On Wed, May 08, 2019 at 10:00:35AM -0700, Florian Fainelli wrote:
> >> If the SCMI firmware implementation is reporting values in a scale that
> >> is different from the HWMON units, we need to scale up or down the value
> >> according to how far appart they are.
> >>
> >> Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
> >> ---
> >>  drivers/hwmon/scmi-hwmon.c | 46 ++++++++++++++++++++++++++++++++++++++
> >>  1 file changed, 46 insertions(+)
> >>
> >> diff --git a/drivers/hwmon/scmi-hwmon.c b/drivers/hwmon/scmi-hwmon.c
> >> index a80183a488c5..4399372e2131 100644
> >> --- a/drivers/hwmon/scmi-hwmon.c
> >> +++ b/drivers/hwmon/scmi-hwmon.c
> >> @@ -7,6 +7,7 @@
> >>   */
> >>  
> >>  #include <linux/hwmon.h>
> >> +#include <linux/limits.h>
> >>  #include <linux/module.h>
> >>  #include <linux/scmi_protocol.h>
> >>  #include <linux/slab.h>
> >> @@ -18,6 +19,47 @@ struct scmi_sensors {
> >>  	const struct scmi_sensor_info **info[hwmon_max];
> >>  };
> >>  
> >> +static inline u64 __pow10(u8 x)
> >> +{
> >> +	u64 r = 1;
> >> +
> >> +	while (x--)
> >> +		r *= 10;
> >> +
> >> +	return r;
> >> +}
> >> +
> >> +static int scmi_hwmon_scale(const struct scmi_sensor_info *sensor, u64 *value)
> >> +{
> >> +	s8 scale = sensor->scale;
> >> +	u64 f;
> >> +
> >> +	switch (sensor->type) {
> >> +	case TEMPERATURE_C:
> >> +	case VOLTAGE:
> >> +	case CURRENT:
> >> +		scale += 3;
> >> +		break;
> >> +	case POWER:
> >> +	case ENERGY:
> >> +		scale += 6;
> >> +		break;
> >> +	default:
> >> +		break;
> >> +	}
> >> +
> >> +	f = __pow10(abs(scale));
> >> +	if (f == U64_MAX)
> >> +		return -E2BIG;
> > 
> > Unfortunately that is not how integer overflows work.
> > 
> > A test program with increasing values of scale reports:
> > 
> > 0: 1
> > ...
> > 18: 1000000000000000000
> > 19: 10000000000000000000
> > 20: 7766279631452241920
> > 21: 3875820019684212736
> > 22: 1864712049423024128
> > 23: 200376420520689664
> > 24: 2003764205206896640
> > ...
> > 61: 11529215046068469760
> > 62: 4611686018427387904
> > 63: 9223372036854775808
> > 64: 0
> > ...
> > 
> > You'll have to check for abs(scale) > 19 if you want to report overflows.
> 
> Yes silly me, my test program was flawed, thanks for pointing out that.
> You are okay with returning E2BIG when we overflow?

Yes.

Thanks,
Guenter