Re: HWMON periodically collection

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Thu, Oct 12, 2023 at 01:18:37PM +0300, Shenhar, Talel wrote:
> 
> On 10/11/2023 6:09 PM, Guenter Roeck wrote:
> > 
> > On Wed, Oct 11, 2023 at 10:45:03AM +0000, Ben Shaul, Almog wrote:
> > > Hi all,
> > > 
> > > We'd like to collect hwmon sensors and get their min/max/avg for our platforms.
> > > 
> > > Those sensors doesn't support min/max by HW and only support reading current values (they also don't support avg but that is likely the case for all other devices).
> > > 
> > > For that goal of getting min/max/avg we have two options:
> > > 
> > > 1. Write userspace service that will constantly read the values and do the calculations (and later report to our database)
> > Yes, this is what you'll have to do.
> We appreciate the prompt response!
> > 
> > > 2. Extend HWMON subsystem/library so each registering hwmon device will ask the subsystem to do it for you.
> > >      Then kernel will create workqueue and constantly read the values and make the calculations (which shall later be readable via hwmon sysfs)
> > No. This would add runtime overhead to each hwmon device even if the
> > information is not used (which would be the case for almost every user).
> 
> That is a fair point.
> 
> What about enabling this during runtime in case user will want to utilize
> this?
> 
> e.g. ioctl/sysfs/debugfs
> 

No.

> Of course users can configure the interval during runtime (or even devfreq
> based?).
> 
> 
> The only concern we had for doing it from usrespace is the short interval
> (and multiple sensors) required will create overhead. Of course doing that
> in kernel space doesn't make this overhead go away and even worsen it as we
> don't go via scheduler. (but does ease userspace/kernelspace context
> switch).

Following your logic, we should implement pretty much everything in the
kernel to avoid userspace/kernelspace context switches. Sorry, we'll have
to agree to disagree here.

> 
> And of course the added value of contributing such a feature which does make
> sense as probably many want to know across their HW what is the power
> consumption. are you aware of some package that supply this kind of info? or
> everyone interested is doing some sort of home made solution?
> 
Presumably various userspace packages implement this, but I don't know of
any specific ones.

Guenter



[Index of Archives]     [LM Sensors]     [Linux Sound]     [ALSA Users]     [ALSA Devel]     [Linux Audio Users]     [Linux Media]     [Kernel]     [Gimp]     [Yosemite News]     [Linux Media]

  Powered by Linux