wpa_supplicant ad-hoc settings

Jouni Malinen j at w1.fi
Thu Nov 18 14:28:30 EST 2010


On Thu, Nov 18, 2010 at 02:50:40PM +0100, Per-Henrik Lundblom wrote:
> The problem with this is that I don't think a WLAN driver should need to 
> have any information about what surrounding networks. Unfortunately that 
> is necessary if the wpa_supplicant doesn't provide provide information 
> about BSSID, ESSID, mode and channel to the driver. Especially the 
> channel is interesting. Of course, one could add the frequency directive 
> to the network configuration but that makes the whole scenario less 
> flexible.

You'll get the channel for the case where a new IBSS is to be created..
SSID is obviously also provided. BSSID generation for IBSS is something
that is expected to be taken care of by the driver in the same way as
merging IBSS STAs that happen to be using different BSSID. Are you
suggesting that wpa_supplicant would take more active role in IBSS BSSID
generation, merges, and channel selection?

> > It would be possible to extend the ap_scan=1 case to handle discovered
> > IBSS, i.e., select to use the IBSS network block in such a case.
> > However, if there are multiple enabled network blocks, it is unclear
> > what should be done if one or more of them are IBSS networks and there
> > are no matching scan results. Only the case of having a single enabled
> > network block would be clear in such a case.
> 
> Yes and no, looking at the current source (0.7.3), after 
> wpa_supplicant_pick_network() didn't find a network wpa_supplicant tries 
> to create a new network. In the function 
> wpa_supplicant_pick_new_network() the list of configured networks are 
> traversed for the highest prio network. This should be valid for ad-hoc 
> networks too.

Which the is the part I was referring to with it being possible to
extend this..

> Or is it just me who thinks it is wrong that it is up to every WLAN 
> driver to handle the ad-hoc case? I know it can be hard to avoid two 
> nodes ending up in their own ad-hoc networks instead of joining each 
> other. Still, I think better support in wpa_supplicant is a key feature.

Depends on what you including in handling IBSS.. wpa_supplicant takes
care of some parts, driver others.. I don't see that ever changing; the
border between the modules may move, but frankly, IBSS is not really
that high on my interest list and I would much rather use my time on
station and AP mode.

-- 
Jouni Malinen                                            PGP id EFC895FA


More information about the HostAP mailing list