lkml.org 
[lkml]   [2022]   [Aug]   [10]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
SubjectRe: [PATCH v2] hwmon: corsair-psu: add reporting of rail mode via debugfs
From
On 8/10/22 10:48, Wilken Gottwalt wrote:
> On Wed, 10 Aug 2022 10:29:08 -0700
> Guenter Roeck <linux@roeck-us.net> wrote:
>
>> On 8/10/22 09:56, Wilken Gottwalt wrote:
>>> On Wed, 10 Aug 2022 09:31:21 -0700
>>> Guenter Roeck <linux@roeck-us.net> wrote:
>>>
>>>> On 8/10/22 06:53, Wilken Gottwalt wrote:
>>>>> Add reporting if the PSU is running in single or multi rail mode via
>>>>> ocpmode debugfs entry. Also update the documentation accordingly.
>>>>>
>>>>> Signed-off-by: Wilken Gottwalt <wilken.gottwalt@posteo.net>
>>>>> ---
>>>>> Changes in v2:
>>>>> - fixed spelling issues in commit message
>>>>
>>>> You did not address or even provide feedback on my second comment.
>>>
>>> Oh darn ... sorry, I was quite busy and didn't really pay attention. I will
>>> answer the earlier mail and think about it.
>>>
>>> Though, maybe you can help me with that what keeps me so busy. Would it be okay
>>> to use a kthread in a hwmon driver to do sampling (500ms - 10s) in conjunction
>>> with HWMON_C_UPDATE_INTERVAL, or is this a strict no-no? I know it is actually
>>> used to set a sample/update rate in a sensor (-register), but this USB-HID
>>> approach is a pure polling thing. It seems to work quite and enables the driver
>>> to collect data quite early in the boot process.
>>>
>>
>> It really depends. Is it _necessary_ ? The pwm-fan driver uses a timer for
>> periodic polling, but that is because it has to. We should not do it purely
>> for convenience, and from the code I don't immediately see why it would
>> be necessary.
>
> Together with the polling I would add encountered lowest and highest values and
> the average of basically all available sensors (kind of session statistics). I
> know it is a bit odd, but currently these power supplies are sold again in a
> newer version and people really like to use them in their servers/workstations
> because of the "realtime" data and this driver. No joke, but I really got
> several requests to add this and I must admit I have quite some fun implementing
> it.
>

That is out of scope for a kernel driver. If desired, a user space application
should do the polling and calculate statistics such as lowest/highest or averages.

Guenter

> Maybe I can provide a patch after I'm done and you can decide if this is okay
> or not. After all I provide an external more enhanced driver via github where
> some features are added, which are a clear no-go. It would be nice to have that
> in mainline, but I'm absolutely fine with a "no".
>
> greetings,
> Wilken

\
 
 \ /
  Last update: 2022-08-10 20:22    [W:0.060 / U:2.568 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site