[PATCH] p2p_supplicant: probe reporting should be from listen interface

[Johannes Berg]

On Fri, 2013-04-05 at 10:12 +0200, Arend van Spriel wrote:

> >>> Found during testing that my device was not found, because we were not
> >>> responding to P2P probe requests during the listen phase.
> >>>
> >>> This patch fixes it for my driver and it kind of makes sense to me
> >>> that the
> >>> probe request reporting should be done on the same interface as the on
> >>> doing the LISTEN.
> >>>
> >>> Not sure whether the P2P device patches are already applied so
> >>> sending this
> >>> patch to you first.
> >>>
> >>
> >> Hmmm.
> >> Your logic seems to be perfect, but it works fine for me without your
> >> fix, and stops to work with it. Didn't debug it yet, but maybe
> >> something is wrong with mac80211?
> >
> > As I am working with a fullmac driver it does not involve mac80211. Are
> > you using a mac80211 driver? Otherwise, we have to look for another
> > culprit ;-)

> It seems we are having an interworking issue here. During the P2P LISTEN 
> state, should probe requests be reported on the P2P_DEVICE interface or 
> on the primary interface? The probe response from wpa_supplicant should 
> be sent on the P2P_DEVICE interface for sure, right? We need a consensus 
> here on what behaviour is required.

I think they should be reported on the P2P_DEVICE. I believe that's
actually the case for our driver anyway since it just reports the
frames, and mac80211 will sort them according to the virtual interface
address.

OTOH, if this change broke our driver and it worked before, maybe it's
not the case? I don't really know.

David, can you send me (privately, and probably better to my @intel
address) a trace capture for this problem?

johannes