On Mon, Feb 18, 2013 at 07:29:47 -0800, Johannes Berg wrote:
> On Mon, 2013-02-18 at 15:46 +0100, Antonio Quartulli wrote:
> 
> > In my current implementation I created a "twin hash-table". It contains
> > statistics for *all* the stations (peer and non-peer).
> > 
> > I think that instead of embedding this new struct (let's call it sta_stats) into
> > the sta_info one, it would be easier to let them be independent (this is why I
> > created the twin hash) and then create
> > a pointer from the sta_info to the related sta_stats.
> 
> I don't really see value in that, it would only make the implementation
> less efficient, because either you follow another pointer (sta->stats)
> or you have to look in the other hash table. That's why I prefer
> embedding it, we have to do the station hash table lookup anyway.

I did not like this approach because the sta_info struct is so big that
when we want to fill the stats substruct only we will waste a lot of bytes.

I think this is the usual space vs. time tread-off :).
However, since this struct is accessed on the critical path, it may be worth to
save time and go for the "embed" approach. Let's see what the other thinks about
it....

> 
> > For the API I think we should create a new nl80211 command.
> > If we simply add a
> > flag to the normal "station dump" command, we would not have all the attributes
> > to print (keep in mind station dump prints attributes that are in sta_info and
> > that are not in sta_stats).
> 
> station dump can just print all attributes that exist, some stations
> would have more/different attributes than other stations. But I don't
> really think it makes a big difference either way, reusing it would be
> fine if the default is to not include the stats-only entries.

yeah, the default behaviour should remain the same (print peer stations
only).


Cheers,

-- 
Antonio Quartulli

..each of us alone is worth nothing..
Ernesto "Che" Guevara