Re: Maximum rate value of input power for pmbus device

Re: Maximum rate value of input power for pmbus device

[Guenter Roeck]

+ linux-hwmon@vger.kernel.org

On Mon, Jun 29, 2020 at 08:31:11AM +0200, zlukwins wrote:
> Hi,
> 
> 
> I am OpenBMC FW developer working currently on some power measurement stuff.
> 
> I would like to have maximum rated input power for pmubus device available
> in hwmon sysfs. This value is read by MFR_PIN_MAX command:
> 
> /MFR_PIN_MAX//
> //The MFR_PIN_MIN command sets or retrieves the maximum rated value, in
> watts, of//
> //the input power./

Interesting typo in the PMBus specification. Yes, it really does associate
MFR_PIN_MIN - which doesn't seem to exist - with the maximum rated output
power.

> 
> And I wondering which attribute shell be used to expose that value in sysfs.
> I went through documentation
> (https://www.kernel.org/doc/Documentation/hwmon/sysfs-interface) and found
> the following:
> 
> power[1-*]_max			Maximum power.
> 				Unit: microWatt
> 				RW
> 
> But it looks like it is already occupied by PIN_OP_WARN_LIMIT.
> 
> Maybe new attribute shall be used? If so how to call that?
> 

None of the standard attributes reports (or is supposed to report) rated
values, so we can not just use any of those.

Also, we can not just add a single attribute to handle this situation,
for the simple reason that there are many more similar attributes.
PMBus specifies (this is from version 1.3.1):

MFR_VIN_MIN
MFR_VIN_MAX
MFR_IIN_MAX
MFR_PIN_MAX
MFR_VOUT_MIN
MFR_VOUT_MAX
MFR_IOUT_MAX
MFR_POUT_MAX
MFR_TAMBIENT_MAX
MFR_TAMBIENT_MIN
MFR_MAX_TEMP_{1,2.3}

All those report rated values. I do see the need/desire for reporting such
information. The only real solution I can see is to add a new set of
attributes to the hwmon ABI. Something like:

	currentX_rated_min	# for consistency
	currentX_rated_max
	inX_rated_min
	inX_rated_max
	powerX_rated_min	# for consistency
	powerX_rated_max
	tempX_rated_min
	tempX_rated_max
plus maybe, for consistency:
	humidityX_rated_min
	humidityX_rated_max

Those would be read-only attributes.

Thoughts, comments, feedback anyone ?

Thanks,
Guenter

Re: Maximum rate value of input power for pmbus device

[zlukwins]

On 6/29/20 6:59 PM, Guenter Roeck wrote:
> +linux-hwmon@vger.kernel.org
>
> On Mon, Jun 29, 2020 at 08:31:11AM +0200, zlukwins wrote:
>> Hi,
>>
>>
>> I am OpenBMC FW developer working currently on some power measurement stuff.
>>
>> I would like to have maximum rated input power for pmubus device available
>> in hwmon sysfs. This value is read by MFR_PIN_MAX command:
>>
>> /MFR_PIN_MAX//
>> //The MFR_PIN_MIN command sets or retrieves the maximum rated value, in
>> watts, of//
>> //the input power./
> Interesting typo in the PMBus specification. Yes, it really does associate
> MFR_PIN_MIN - which doesn't seem to exist - with the maximum rated output
> power.
>> And I wondering which attribute shell be used to expose that value in sysfs.
>> I went through documentation
>> (https://www.kernel.org/doc/Documentation/hwmon/sysfs-interface) and found
>> the following:
>>
>> power[1-*]_max			Maximum power.
>> 				Unit: microWatt
>> 				RW
>>
>> But it looks like it is already occupied by PIN_OP_WARN_LIMIT.
>>
>> Maybe new attribute shall be used? If so how to call that?
>>
> None of the standard attributes reports (or is supposed to report) rated
> values, so we can not just use any of those.
>
> Also, we can not just add a single attribute to handle this situation,
> for the simple reason that there are many more similar attributes.
> PMBus specifies (this is from version 1.3.1):
>
> MFR_VIN_MIN
> MFR_VIN_MAX
> MFR_IIN_MAX
> MFR_PIN_MAX
> MFR_VOUT_MIN
> MFR_VOUT_MAX
> MFR_IOUT_MAX
> MFR_POUT_MAX
> MFR_TAMBIENT_MAX
> MFR_TAMBIENT_MIN
> MFR_MAX_TEMP_{1,2.3}
>
> All those report rated values. I do see the need/desire for reporting such
> information. The only real solution I can see is to add a new set of
> attributes to the hwmon ABI. Something like:
>
> 	currentX_rated_min	# for consistency
> 	currentX_rated_max
> 	inX_rated_min
> 	inX_rated_max
> 	powerX_rated_min	# for consistency
> 	powerX_rated_max
> 	tempX_rated_min
> 	tempX_rated_max
> plus maybe, for consistency:
> 	humidityX_rated_min
> 	humidityX_rated_max
>
> Those would be read-only attributes.
>
> Thoughts, comments, feedback anyone ?
>
> Thanks,
> Guenter


I really like your proposition but I guess we need to wait few more days 
for the feedback.

But have some questions here.

What about potential next steps when we all agreed to follow that 
approach. Should documentation modification reach upstream repository 
first and then e.g. pmbus hwmon module implementation?


Thanks

Zbigniew

Re: Maximum rate value of input power for pmbus device

[Guenter Roeck]

On 7/8/20 3:50 AM, zlukwins wrote:
> 
> On 6/29/20 6:59 PM, Guenter Roeck wrote:
>> +linux-hwmon@vger.kernel.org
>>
>> On Mon, Jun 29, 2020 at 08:31:11AM +0200, zlukwins wrote:
>>> Hi,
>>>
>>>
>>> I am OpenBMC FW developer working currently on some power measurement stuff.
>>>
>>> I would like to have maximum rated input power for pmubus device available
>>> in hwmon sysfs. This value is read by MFR_PIN_MAX command:
>>>
>>> /MFR_PIN_MAX//
>>> //The MFR_PIN_MIN command sets or retrieves the maximum rated value, in
>>> watts, of//
>>> //the input power./
>> Interesting typo in the PMBus specification. Yes, it really does associate
>> MFR_PIN_MIN - which doesn't seem to exist - with the maximum rated output
>> power.
>>> And I wondering which attribute shell be used to expose that value in sysfs.
>>> I went through documentation
>>> (https://www.kernel.org/doc/Documentation/hwmon/sysfs-interface) and found
>>> the following:
>>>
>>> power[1-*]_max            Maximum power.
>>>                 Unit: microWatt
>>>                 RW
>>>
>>> But it looks like it is already occupied by PIN_OP_WARN_LIMIT.
>>>
>>> Maybe new attribute shall be used? If so how to call that?
>>>
>> None of the standard attributes reports (or is supposed to report) rated
>> values, so we can not just use any of those.
>>
>> Also, we can not just add a single attribute to handle this situation,
>> for the simple reason that there are many more similar attributes.
>> PMBus specifies (this is from version 1.3.1):
>>
>> MFR_VIN_MIN
>> MFR_VIN_MAX
>> MFR_IIN_MAX
>> MFR_PIN_MAX
>> MFR_VOUT_MIN
>> MFR_VOUT_MAX
>> MFR_IOUT_MAX
>> MFR_POUT_MAX
>> MFR_TAMBIENT_MAX
>> MFR_TAMBIENT_MIN
>> MFR_MAX_TEMP_{1,2.3}
>>
>> All those report rated values. I do see the need/desire for reporting such
>> information. The only real solution I can see is to add a new set of
>> attributes to the hwmon ABI. Something like:
>>
>>     currentX_rated_min    # for consistency
>>     currentX_rated_max
>>     inX_rated_min
>>     inX_rated_max
>>     powerX_rated_min    # for consistency
>>     powerX_rated_max
>>     tempX_rated_min
>>     tempX_rated_max
>> plus maybe, for consistency:
>>     humidityX_rated_min
>>     humidityX_rated_max
>>
>> Those would be read-only attributes.
>>
>> Thoughts, comments, feedback anyone ?
>>
>> Thanks,
>> Guenter
> 
> 
> I really like your proposition but I guess we need to wait few more days for the feedback.
> 
> But have some questions here.
> 
> What about potential next steps when we all agreed to follow that approach. Should documentation modification reach upstream repository first and then e.g. pmbus hwmon module implementation?
> 
We'll need a series of patches. One to amend the documentation, one to add
the necessary code to the hwmon core (so that the core supports it with
the _with_info API), one to add support to the PMBus core, and one each
to add support to affected drivers.

Once this is all complete, the lm-sensors package should be updated
as well.

Guenter

Re: Maximum rate value of input power for pmbus device

[zlukwins]

On 7/8/20 3:33 PM, Guenter Roeck wrote:
> On 7/8/20 3:50 AM, zlukwins wrote:
>> On 6/29/20 6:59 PM, Guenter Roeck wrote:
>>> +linux-hwmon@vger.kernel.org
>>>
>>> On Mon, Jun 29, 2020 at 08:31:11AM +0200, zlukwins wrote:
>>>> Hi,
>>>>
>>>>
>>>> I am OpenBMC FW developer working currently on some power measurement stuff.
>>>>
>>>> I would like to have maximum rated input power for pmubus device available
>>>> in hwmon sysfs. This value is read by MFR_PIN_MAX command:
>>>>
>>>> /MFR_PIN_MAX//
>>>> //The MFR_PIN_MIN command sets or retrieves the maximum rated value, in
>>>> watts, of//
>>>> //the input power./
>>> Interesting typo in the PMBus specification. Yes, it really does associate
>>> MFR_PIN_MIN - which doesn't seem to exist - with the maximum rated output
>>> power.
>>>> And I wondering which attribute shell be used to expose that value in sysfs.
>>>> I went through documentation
>>>> (https://www.kernel.org/doc/Documentation/hwmon/sysfs-interface) and found
>>>> the following:
>>>>
>>>> power[1-*]_max            Maximum power.
>>>>                  Unit: microWatt
>>>>                  RW
>>>>
>>>> But it looks like it is already occupied by PIN_OP_WARN_LIMIT.
>>>>
>>>> Maybe new attribute shall be used? If so how to call that?
>>>>
>>> None of the standard attributes reports (or is supposed to report) rated
>>> values, so we can not just use any of those.
>>>
>>> Also, we can not just add a single attribute to handle this situation,
>>> for the simple reason that there are many more similar attributes.
>>> PMBus specifies (this is from version 1.3.1):
>>>
>>> MFR_VIN_MIN
>>> MFR_VIN_MAX
>>> MFR_IIN_MAX
>>> MFR_PIN_MAX
>>> MFR_VOUT_MIN
>>> MFR_VOUT_MAX
>>> MFR_IOUT_MAX
>>> MFR_POUT_MAX
>>> MFR_TAMBIENT_MAX
>>> MFR_TAMBIENT_MIN
>>> MFR_MAX_TEMP_{1,2.3}
>>>
>>> All those report rated values. I do see the need/desire for reporting such
>>> information. The only real solution I can see is to add a new set of
>>> attributes to the hwmon ABI. Something like:
>>>
>>>      currentX_rated_min    # for consistency
>>>      currentX_rated_max
>>>      inX_rated_min
>>>      inX_rated_max
>>>      powerX_rated_min    # for consistency
>>>      powerX_rated_max
>>>      tempX_rated_min
>>>      tempX_rated_max
>>> plus maybe, for consistency:
>>>      humidityX_rated_min
>>>      humidityX_rated_max
>>>
>>> Those would be read-only attributes.
>>>
>>> Thoughts, comments, feedback anyone ?
>>>
>>> Thanks,
>>> Guenter
>>
>> I really like your proposition but I guess we need to wait few more days for the feedback.
>>
>> But have some questions here.
>>
>> What about potential next steps when we all agreed to follow that approach. Should documentation modification reach upstream repository first and then e.g. pmbus hwmon module implementation?
>>
> We'll need a series of patches. One to amend the documentation, one to add
> the necessary code to the hwmon core (so that the core supports it with
> the _with_info API), one to add support to the PMBus core, and one each
> to add support to affected drivers.
>
> Once this is all complete, the lm-sensors package should be updated
> as well.
>
> Guenter

Do you think we need to wait for feedback or maybe we could implement 
with your proposition? Maybe I could start working on patches?


Thanks

Zbigniew

Re: Maximum rate value of input power for pmbus device

[Guenter Roeck]

On 7/21/20 11:31 AM, zlukwins wrote:
> 
> On 7/8/20 3:33 PM, Guenter Roeck wrote:
>> On 7/8/20 3:50 AM, zlukwins wrote:
>>> On 6/29/20 6:59 PM, Guenter Roeck wrote:
>>>> +linux-hwmon@vger.kernel.org
>>>>
>>>> On Mon, Jun 29, 2020 at 08:31:11AM +0200, zlukwins wrote:
>>>>> Hi,
>>>>>
>>>>>
>>>>> I am OpenBMC FW developer working currently on some power measurement stuff.
>>>>>
>>>>> I would like to have maximum rated input power for pmubus device available
>>>>> in hwmon sysfs. This value is read by MFR_PIN_MAX command:
>>>>>
>>>>> /MFR_PIN_MAX//
>>>>> //The MFR_PIN_MIN command sets or retrieves the maximum rated value, in
>>>>> watts, of//
>>>>> //the input power./
>>>> Interesting typo in the PMBus specification. Yes, it really does associate
>>>> MFR_PIN_MIN - which doesn't seem to exist - with the maximum rated output
>>>> power.
>>>>> And I wondering which attribute shell be used to expose that value in sysfs.
>>>>> I went through documentation
>>>>> (https://www.kernel.org/doc/Documentation/hwmon/sysfs-interface) and found
>>>>> the following:
>>>>>
>>>>> power[1-*]_max            Maximum power.
>>>>>                  Unit: microWatt
>>>>>                  RW
>>>>>
>>>>> But it looks like it is already occupied by PIN_OP_WARN_LIMIT.
>>>>>
>>>>> Maybe new attribute shall be used? If so how to call that?
>>>>>
>>>> None of the standard attributes reports (or is supposed to report) rated
>>>> values, so we can not just use any of those.
>>>>
>>>> Also, we can not just add a single attribute to handle this situation,
>>>> for the simple reason that there are many more similar attributes.
>>>> PMBus specifies (this is from version 1.3.1):
>>>>
>>>> MFR_VIN_MIN
>>>> MFR_VIN_MAX
>>>> MFR_IIN_MAX
>>>> MFR_PIN_MAX
>>>> MFR_VOUT_MIN
>>>> MFR_VOUT_MAX
>>>> MFR_IOUT_MAX
>>>> MFR_POUT_MAX
>>>> MFR_TAMBIENT_MAX
>>>> MFR_TAMBIENT_MIN
>>>> MFR_MAX_TEMP_{1,2.3}
>>>>
>>>> All those report rated values. I do see the need/desire for reporting such
>>>> information. The only real solution I can see is to add a new set of
>>>> attributes to the hwmon ABI. Something like:
>>>>
>>>>      currentX_rated_min    # for consistency
>>>>      currentX_rated_max
>>>>      inX_rated_min
>>>>      inX_rated_max
>>>>      powerX_rated_min    # for consistency
>>>>      powerX_rated_max
>>>>      tempX_rated_min
>>>>      tempX_rated_max
>>>> plus maybe, for consistency:
>>>>      humidityX_rated_min
>>>>      humidityX_rated_max
>>>>
>>>> Those would be read-only attributes.
>>>>
>>>> Thoughts, comments, feedback anyone ?
>>>>
>>>> Thanks,
>>>> Guenter
>>>
>>> I really like your proposition but I guess we need to wait few more days for the feedback.
>>>
>>> But have some questions here.
>>>
>>> What about potential next steps when we all agreed to follow that approach. Should documentation modification reach upstream repository first and then e.g. pmbus hwmon module implementation?
>>>
>> We'll need a series of patches. One to amend the documentation, one to add
>> the necessary code to the hwmon core (so that the core supports it with
>> the _with_info API), one to add support to the PMBus core, and one each
>> to add support to affected drivers.
>>
>> Once this is all complete, the lm-sensors package should be updated
>> as well.
>>
>> Guenter
> 
> Do you think we need to wait for feedback or maybe we could implement with your proposition? Maybe I could start working on patches?
> 

I don't think we will get any further feedback. Please go ahead.

Thanks,
Guenter

> 
> Thanks
> 
> Zbigniew
>