Re: [PATCH 1/2] netdev: Add logging for CQM messages

[Denis Kenzior <denkenz@gmail.com>]

Hi Michael,

On 8/1/22 13:09, Michael Johnson wrote:
> Hi Denis,
> 
> On Mon, 1 Aug 2022 at 18:24, Denis Kenzior <denkenz@gmail.com> wrote:
>> I wonder if we should report this over D-Bus instead?  I mean you can always
>> setup your own SignalLevelAgent, or we could even add more reporting via
>> StationDiagnostic interface?
> 
> I actually do record the signal level separately but it's stored in an
> annoying binary format so access is more painful :(. This was just to
> make things easier and reduce the jump from wpa_supplicant to iwd for
> anyone on plain ubuntu etc. I can understand why you don't want the
> spam though.

I don't mind the spam in debug mode, that's what it is for.  But not sure they 
really belong at info/warn levels.  I think we can consider putting in roaming 
related messages at the info level since that type of information might be 
generally useful.

> 
>> Right, but I think it sets a single CQM threshold, not multiple like we (might)
>> do.  Also, wpa_s is very spammy since it is really (only) meant for testing wifi
>> implementations in a lab ;)
> 
> haha fair. Of course, it's still the default on most linux systems ;).
> 

And we're working to change that.

> 
> What would you like to do with the patch? I'm happy to drop it if its
> too spammy or we can lower the level to debug? Or make it only print
> when the threshold changes.

I don't mind these at debug level, if you want to start with that.

> Are you happy with the packet loss and beacon loss events or are they
> too much as well?

As I mentioned before, kernel will generally disconnect us as soon as the beacon 
loss event happens (depends on the STA capability if I recall correctly).  Not 
sure how really useful this is, but since beacon loss goes via dmesg at 'info' 
level, I think it would be fine for iwd to do so as well.

The packet loss one I'm not sure.  Certaintly not 'warn'.  I think I'd start 
with 'debug' since this event isn't being reacted to (yet).

Regards,
-Denis