phosphor-host-ipmid sensor SDRs on dbus

phosphor-host-ipmid sensor SDRs on dbus

[Patrick Venture]

I sent out an email indicating that we need to make sensors somewhat
dynamic in phosphor-host-ipmid, and either using json or dbus was a
reasonable approach.  I think json would be easier, but, for those
using entity-manager, perhaps having the SDR information supported in
some way associated with the dbus output might be useful.

Idea: Given a sensor, read the SDR from dbus.
Problem: association of a sensor id with a sensor on dbus -- maybe
some search step is run once that caches everything and listens for
change (similar to what James did with phosphor-pid-control)...

I'm just throwing out ideas without full design changes because I want
to get a sense of where the preferences lie before really digging into
what changes are required.

It would support both types (like dbus configuration in phosphor-pid-control).

Patrick

Re: phosphor-host-ipmid sensor SDRs on dbus

[James Feist]

On 7/23/2019 1:40 PM, Patrick Venture wrote:
> I sent out an email indicating that we need to make sensors somewhat
> dynamic in phosphor-host-ipmid, and either using json or dbus was a
> reasonable approach.  I think json would be easier, but, for those
> using entity-manager, perhaps having the SDR information supported in
> some way associated with the dbus output might be useful.

In intel-ipmi-oem we were able to use the max / min values to generate 
an SDR. The only real hold ups for global adoption I believe were:
1. static sensor numbers were needed.
2. it was implemented using doubles for sensor numbers

If you had some configuration so that you could make these things 
non-generated if you wished, this might be useful to you.

https://github.com/openbmc/intel-ipmi-oem/blob/edf25e614919d0b379b885384bcdac987580b844/src/sensorcommands.cpp#L1259

-James

> 
> Idea: Given a sensor, read the SDR from dbus.
> Problem: association of a sensor id with a sensor on dbus -- maybe
> some search step is run once that caches everything and listens for
> change (similar to what James did with phosphor-pid-control)...
> 
> I'm just throwing out ideas without full design changes because I want
> to get a sense of where the preferences lie before really digging into
> what changes are required.
> 
> It would support both types (like dbus configuration in phosphor-pid-control).
> 
> Patrick
>