i5k_amb temp_mid replacement

i5k_amb temp_mid replacement

[Armin Wolf]

Hello,

while trying to convert i5k_temp to the new hwmon API to resolve
https://bugzilla.kernel.org/show_bug.cgi?id=208693,
i was asking myself whether or not temp_mid could be replaced with a
standard sysfs attribute?
Since afaik temp_mid and temp_max are both temperature limits, they
could potentially be replaced with
temp_max and temp_crit.

Armin Wolf.

Re: i5k_amb temp_mid replacement

[Guenter Roeck]

On Fri, Oct 01, 2021 at 12:13:53AM +0200, Armin Wolf wrote:
> Hello,
> 
> while trying to convert i5k_temp to the new hwmon API to resolve
> https://bugzilla.kernel.org/show_bug.cgi?id=208693,
> i was asking myself whether or not temp_mid could be replaced with a
> standard sysfs attribute?
> Since afaik temp_mid and temp_max are both temperature limits, they
> could potentially be replaced with
> temp_max and temp_crit.
> 

Quite likely. Unfortunately, the chipset documentation is not available
to the public, so it is difficult to determine what those temperatures
actually mean.

Guenter

Re: i5k_amb temp_mid replacement

[Armin Wolf]

Am 02.10.21 um 16:04 schrieb Guenter Roeck:
> On Fri, Oct 01, 2021 at 12:13:53AM +0200, Armin Wolf wrote:
>> Hello,
>>
>> while trying to convert i5k_temp to the new hwmon API to resolve
>> https://bugzilla.kernel.org/show_bug.cgi?id=208693,
>> i was asking myself whether or not temp_mid could be replaced with a
>> standard sysfs attribute?
>> Since afaik temp_mid and temp_max are both temperature limits, they
>> could potentially be replaced with
>> temp_max and temp_crit.
>>
> Quite likely. Unfortunately, the chipset documentation is not available
> to the public, so it is difficult to determine what those temperatures
> actually mean.
>
> Guenter
Darrick Wong says the chipset documentation would support such a change.
However, he fears that the changed meaning of tempX_max could confuse
userspace
programs.
But i think the current not using standard attribute names will confuse
even more
programms.

Re: i5k_amb temp_mid replacement

[Guenter Roeck]

On 10/7/21 7:20 AM, Armin Wolf wrote:
> Am 02.10.21 um 16:04 schrieb Guenter Roeck:
>> On Fri, Oct 01, 2021 at 12:13:53AM +0200, Armin Wolf wrote:
>>> Hello,
>>>
>>> while trying to convert i5k_temp to the new hwmon API to resolve
>>> https://bugzilla.kernel.org/show_bug.cgi?id=208693,
>>> i was asking myself whether or not temp_mid could be replaced with a
>>> standard sysfs attribute?
>>> Since afaik temp_mid and temp_max are both temperature limits, they
>>> could potentially be replaced with
>>> temp_max and temp_crit.
>>>
>> Quite likely. Unfortunately, the chipset documentation is not available
>> to the public, so it is difficult to determine what those temperatures
>> actually mean.
>>
>> Guenter
> Darrick Wong says the chipset documentation would support such a change.
> However, he fears that the changed meaning of tempX_max could confuse
> userspace
> programs.
> But i think the current not using standard attribute names will confuse
> even more
> programms.

I still don't know what temperature limits temp_mid and temp_max actually
reflect, so I still can not really comment.

Guenter

Re: i5k_amb temp_mid replacement

[Armin Wolf]

Am 07.10.21 um 18:30 schrieb Guenter Roeck:
> On 10/7/21 7:20 AM, Armin Wolf wrote:
>> Am 02.10.21 um 16:04 schrieb Guenter Roeck:
>>> On Fri, Oct 01, 2021 at 12:13:53AM +0200, Armin Wolf wrote:
>>>> Hello,
>>>>
>>>> while trying to convert i5k_temp to the new hwmon API to resolve
>>>> https://bugzilla.kernel.org/show_bug.cgi?id=208693,
>>>> i was asking myself whether or not temp_mid could be replaced with a
>>>> standard sysfs attribute?
>>>> Since afaik temp_mid and temp_max are both temperature limits, they
>>>> could potentially be replaced with
>>>> temp_max and temp_crit.
>>>>
>>> Quite likely. Unfortunately, the chipset documentation is not available
>>> to the public, so it is difficult to determine what those temperatures
>>> actually mean.
>>>
>>> Guenter
>> Darrick Wong says the chipset documentation would support such a change.
>> However, he fears that the changed meaning of tempX_max could confuse
>> userspace
>> programs.
>> But i think the current not using standard attribute names will confuse
>> even more
>> programms.
>
> I still don't know what temperature limits temp_mid and temp_max actually
> reflect, so I still can not really comment.
>
> Guenter
"Intel 6400/6402 Advanced Memory Buffer" seems to document that.

Re: i5k_amb temp_mid replacement

[Guenter Roeck]

On 10/7/21 10:59 AM, Armin Wolf wrote:
> Am 07.10.21 um 18:30 schrieb Guenter Roeck:
>> On 10/7/21 7:20 AM, Armin Wolf wrote:
>>> Am 02.10.21 um 16:04 schrieb Guenter Roeck:
>>>> On Fri, Oct 01, 2021 at 12:13:53AM +0200, Armin Wolf wrote:
>>>>> Hello,
>>>>>
>>>>> while trying to convert i5k_temp to the new hwmon API to resolve
>>>>> https://bugzilla.kernel.org/show_bug.cgi?id=208693,
>>>>> i was asking myself whether or not temp_mid could be replaced with a
>>>>> standard sysfs attribute?
>>>>> Since afaik temp_mid and temp_max are both temperature limits, they
>>>>> could potentially be replaced with
>>>>> temp_max and temp_crit.
>>>>>
>>>> Quite likely. Unfortunately, the chipset documentation is not available
>>>> to the public, so it is difficult to determine what those temperatures
>>>> actually mean.
>>>>
>>>> Guenter
>>> Darrick Wong says the chipset documentation would support such a change.
>>> However, he fears that the changed meaning of tempX_max could confuse
>>> userspace
>>> programs.
>>> But i think the current not using standard attribute names will confuse
>>> even more
>>> programms.
>>
>> I still don't know what temperature limits temp_mid and temp_max actually
>> reflect, so I still can not really comment.
>>
>> Guenter
> "Intel 6400/6402 Advanced Memory Buffer" seems to document that.

Ah, thanks.

 From the description, it should be:

TEMPLO -> temp1_max
TEMPMID -> temp1_crit
TEMPHI -> temp1_emergency

The respective _alarm attributes should be supported as well (via TEMPSTAT),
but that would be a separate patch.

Guenter

Re: i5k_amb temp_mid replacement

[Guenter Roeck]

On 10/7/21 12:24 PM, Guenter Roeck wrote:
> On 10/7/21 10:59 AM, Armin Wolf wrote:
>> Am 07.10.21 um 18:30 schrieb Guenter Roeck:
>>> On 10/7/21 7:20 AM, Armin Wolf wrote:
>>>> Am 02.10.21 um 16:04 schrieb Guenter Roeck:
>>>>> On Fri, Oct 01, 2021 at 12:13:53AM +0200, Armin Wolf wrote:
>>>>>> Hello,
>>>>>>
>>>>>> while trying to convert i5k_temp to the new hwmon API to resolve
>>>>>> https://bugzilla.kernel.org/show_bug.cgi?id=208693,
>>>>>> i was asking myself whether or not temp_mid could be replaced with a
>>>>>> standard sysfs attribute?
>>>>>> Since afaik temp_mid and temp_max are both temperature limits, they
>>>>>> could potentially be replaced with
>>>>>> temp_max and temp_crit.
>>>>>>
>>>>> Quite likely. Unfortunately, the chipset documentation is not available
>>>>> to the public, so it is difficult to determine what those temperatures
>>>>> actually mean.
>>>>>
>>>>> Guenter
>>>> Darrick Wong says the chipset documentation would support such a change.
>>>> However, he fears that the changed meaning of tempX_max could confuse
>>>> userspace
>>>> programs.
>>>> But i think the current not using standard attribute names will confuse
>>>> even more
>>>> programms.
>>>
>>> I still don't know what temperature limits temp_mid and temp_max actually
>>> reflect, so I still can not really comment.
>>>
>>> Guenter
>> "Intel 6400/6402 Advanced Memory Buffer" seems to document that.
> 
> Ah, thanks.
> 
>  From the description, it should be:
> 
> TEMPLO -> temp1_max
> TEMPMID -> temp1_crit
> TEMPHI -> temp1_emergency
> 

Scratch that. I am misreading the document.

TEMPLO -> temp1_min
TEMPMID -> temp1_max
TEMPHI -> temp1_crit

is fine.

Guenter