[enhancement] respondd.statistics.wireless: transmit radioidentifier

[AiyionPrime]

The current implementation of respondd.statistics.wireless does provide a structure much like this:

 "wireless": [
    {
      "frequency": 2412,
      "noise": -95,
      "active": 438652660,
      "busy": 149880419,
      "rx": 136320927,
      "tx": 5144566
    },
    {
      "frequency": 5220,
      "noise": -109,
      "active": 421788737,
      "busy": 12318006,
      "rx": 1747746,
      "tx": 6160088
    }
  ],

Originally posted by @AiyionPrime in #2201 (comment)

This works okayish for single- or dualband devices, as one can distinguish the radios per frequency.
For devices like the ocedo panda or the potentially the turris mox (thanks @blocktrron and @yanosz )
this approach might cause confusion due to their amounts of (possible) radios.

I'd suggest to add a radio-key to the wireless dictionary,
which would hold the corresponding descriptor used in uci context for the wifi-device.

An entry might then look like this:

    {
      "frequency": 5220,
      "noise": -109,
      "active": 421788737,
      "busy": 12318006,
      "idx": 0,
      "rx": 1747746,
      "tx": 6160088
    },

This is heavily related to #644 and a necessity for a potential enhancement in #2201.

[mweinelt]

That would be the way of least resistance, the proper way would be to make it a map. For backwards compatibility I think your proposal sounds good to me.

[lemoer]

I would also prefer keeping backwards compatibility. As long as there is no necessity for breaking changes, let us avoid them.

[blocktrron]

which would hold the corresponding descriptor used in uci context for the wifi-device.

Not a big fan of that. The radioX sections are enumerated in the orer in which the radios have been detected by the system on first boot. This is not necessarily the case when firmware gets updated (happened on the TL-WDR4900 some time ago).

I'd much rather simply transmit the PHY idx of the radio in the mac80211 subsystem.

[AiyionPrime]

til, they're not the same -.-'
Thanks, that was my intention.

[AiyionPrime]

As they're not the same;
I'd like to rename the proposed field idx for clarity.

[AiyionPrime]

PR in packages is open for feedback.

[AiyionPrime]

b72588a closed this.

[neocturne]

I don't like this at all. We have added an arbitrary numeric identifier to the respondd data - that's something we usually avoid in favor of stable identifiers like MAC addresses. How is it used, is there some other data structure that provides the same indices for matching? The array entries already contain the radio's frequency, why is it even relevant which radio it belongs to?

[blocktrron]

@NeoRaider The only other identifier providing the same identification would be the PHY path from the radios UCI options (which i like even less). MAC addresses only identify the VIF, but not the PHY (although you can resolve the PHY).

[neocturne]

So my question remains, what is this data used for, and what is it matched against? If airtime information is most useful for the mesh interfaces, we could use the MAC address of the mesh VIF. Or we could decide that no identifier is necessary at all, because there is nothing it could be matched against anyways.

[mweinelt]

Basically we want to show on the status page which radio is on which frequency, primarily because the frequency is not a per vif thing.

[neocturne]

Basically we want to show on the status page which radio is on which frequency, primarily because the frequency is not a per vif thing.

Hmm, I see - it seems I'm missing the bigger picture here. What information should be displayed on the status page? All info provided by the airtime module?

[neocturne]

One option would be to use the PHY MAC address (/sys/class/ieee80211/phy*/macaddress), although we would also need to add that address to other data structures we want to match the PHY against.

[neocturne]

I'm going to revise the status page code so that the channel display doesn't use respondd data at all (requiring the airtime module for that purpose seems like overkill to me anyways...). The PHY index can then just be removed from the respondd provider again.

[AiyionPrime]

can then just be removed from the respondd provider again.

In that case you might want to engage #644 and update the negative progress of it.