[PATCH] ath10k: handle cycle count wrap around

[Srinivasa Duvvuri]

Please find my responses inline.



>
> For foreign-channel visits this can be easily calculated since they
> are never longer than the wraparound time.

As explained before, the counter values keep increasing across all
channel changes
and they never reset to 0. if they had been reset to zero on every
foreign channel change
then we do not need to handle wrap around as you suggested.
Since they are not reset, we need to handle wrap around .
>
> BSS-channel visits (or between-scans if you will) should be just
> discarded though as you suggested. I'm not fond of re-using
> ar->survey_last_cycle_count though because it always stores a value
> and 0 is just one of them. Adding a special meaning to 0 looks ugly to
> me. I would suggest a dedicated var which tracks whether chan_info
> event without _FLAG_COMPLETE came in. Something along the logic of:
>
>  event_chan_info():
>   ...
>   if (flags & COMPLETE) {
>      if (ar->ch_info_can_process)
>         update(ar->survey);
>
>      ar->ch_info_can_process = false;
>   } else {
>      ar->ch_info_can_process = true;
>   }
>
> This cleanly enforces the expectation that was assumed before, i.e.
> that ch_info comes in pairs always.

agreed this is much cleaner. I will make the change as you suggested.

>
>
> Michał